The strategic stakes are considerable. The geostationary belt hosts a large number of missile early warning and military communication satellites, making it a core strategic resource in orbital operations. In recent years, the United States Geosynchronous Space Situational Awareness Program, known as GSSAP, has conducted hundreds of close-range observation operations on dozens of on-orbit satellites in the vicinity of GEO. These activities have underscored that acquiring multi-aspect information from non-cooperative targets is now a critical component of space security assessment. By observing a target from multiple geometries and ranges, analysts can infer its functions and mission status, turning raw imagery and payload signatures into actionable assessments of what an unknown or suspicious spacecraft is actually built to do.

Close-range spacecraft operations provide the technical means for such observation. Using relative motion configurations such as fly-around, hovering, and drifting flight, an observer can approach and circle a target to collect multi-aspect payload information. Yet existing research has predominantly focused on the control design of a single relative motion configuration, which cannot satisfy the complex requirements of a full multi-aspect observation campaign. Combining multiple observation models introduces a cluster of intertwined challenges: achieving smooth switching between modes, optimizing fuel consumption across the whole sequence, designing tractable strategy algorithms, and analyzing how payload parameters such as camera field of view affect the quality of the resulting data. The new work tackles all of these problems within a single unified framework.

The mathematical foundation of the study rests on the Clohessy–Wiltshire relative dynamics equations, the classical linearized description of relative motion between two spacecraft in near-circular orbits. From the analytical solutions of unforced relative motion, the researchers constructed four typical close-range operation models: the droplet model, the coplanar fly-around model, the non-coplanar fly-around model, and the drifting flight model. For each configuration, the team derived solution functions for the initial motion states that produce stable relative motion trajectories, parameterized by relative distance and phase angle. The droplet model, characterized by its symmetry and revisit capability, enables fine, repeated observation of the target. The coplanar and non-coplanar fly-around models facilitate circling observations in different orbital planes, while the drifting flight model achieves close-range observation through multi-impulse control.

Switching cleanly between these configurations is a nontrivial problem in its own right, because an abrupt transition would waste propellant and could destabilize the relative trajectory. To solve it, the researchers designed a waypoint trajectory planning scheme based on minimum integral squared control theory, a formulation that seeks the control input minimizing the accumulated squared control effort over the transfer. Combined with a multi-impulse maneuver strategy, this scheme allows the observing spacecraft to move smoothly from the terminal state of one observation model to the initial state of the next, stitching the four configurations into a continuous, fuel-conscious observation campaign rather than a set of disconnected maneuvers.

The most novel element of the study is its introduction of sequential coalition game theory into the multi-model combined observation problem, an application the authors describe as the first of its kind. By constructing a game tree whose branches represent the available initial motion points and close-range operation models at each stage, the algorithm searches for the optimal combination at every decision point under minimum fuel constraints. The result is an optimal observation strategy sequence: a staged plan that tells the observing spacecraft which configuration to adopt, where to begin it, and how to transition to the next, all while keeping total propellant expenditure as low as possible. Framing the problem as a sequential game transforms an otherwise combinatorial trajectory planning task into a structured search over well-defined strategic choices.

Simulation results validate the framework end to end. The generated strategy set shows a clear staged progression: the first stage selects an initial point and the droplet model, the second stage selects the drifting flight model, and the third and fourth stages sequentially select the coplanar and non-coplanar fly-around models. The three-dimensional relative trajectories of this four-model combined observation confirm that the algorithm can successfully produce an optimal combined sequence under minimum fuel constraints. In other words, the game-theoretic planner does not merely find a feasible path through the space of possible maneuvers; it finds one that respects the hard propellant budget that ultimately governs how long an observing satellite can remain useful on orbit.

Beyond trajectory planning, the study delivers a quantitative analysis of effective observation time, defined as the total duration during which the target remains observable within prescribed constraints on observation distance and camera field-of-view angle. Because the observation cameras are fixedly installed on the spacecraft, frequent large-angle attitude maneuvers would be both time-consuming and fuel-intensive, making the choice of relative motion geometry the primary lever for maximizing observability. The statistical analysis reveals two practical design rules. First, under the same close-range operation model configuration, increasing the observation distance and the field-of-view angle significantly enhances target observability and yields longer effective observation time. Second, for fixed distance and field of view, reducing the model configuration size also improves observation performance.

These findings carry substantial engineering reference value for space security assessment and on-orbit situational awareness. For mission planners, the work offers a systematic, computable method to design multi-aspect information acquisition campaigns against non-cooperative targets, replacing ad hoc maneuver selection with a principled game-theoretic optimization. For the broader space community, the study arrives at a moment when close-range operations near GEO are no longer hypothetical but routine, and when the ability to characterize an unknown object quickly and cheaply may determine how confidently operators can protect their own assets. By unifying relative orbital dynamics, minimum-fuel trajectory planning, and sequential game theory into one framework, the Sun Yat-sen University team has outlined a template for the next generation of autonomous inspection missions, in which a single observer can adaptively choose how, where, and when to look at a target that will never cooperate.

Subject of Research: Game-theoretic strategy design for multi-model spacecraft proximity observation of non-cooperative satellites

Article Title: Spacecraft proximity operation model-based sequential coalitional observation game strategy design

Article References: Spacecraft proximity operation model-based sequential coalitional observation game strategy design. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: space situational awareness, geostationary orbit, spacecraft proximity operations, sequential coalition game theory, Clohessy–Wiltshire dynamics, non-cooperative targets, fuel optimization, fly-around maneuvers, trajectory planning, GSSAP, space security, relative orbital motion

The system, called SANDO — short for Safe AutoNomous trajectory planning for Dynamic unknOwn environments — was developed by Kota Kondo, who recently completed his doctorate in aeronautics and astronautics at MIT, together with Jesús Tordesillas of Comillas Pontifical University in Madrid, MIT graduate students Juan Rached, Lili Sun, and Yixuan Jia, and senior author Jonathan P. How, the Ford Professor of Aeronautics and Astronautics at MIT. The work appears in the journal IEEE Transactions on Robotics. Its central claim is striking in its simplicity: the only prior knowledge the planner requires is the maximum speed an obstacle could possibly reach. Given that single number, the algorithm can chart a course through a completely unmapped, dynamic environment and prove, before the drone ever takes off, that the resulting trajectory will remain free of collisions.

“In the hardest possible environment, where the UAV has no map of the area and there are unknown obstacles moving around, we established a mathematical guarantee of safety,” Kondo says. “The only thing the planner needs to know is the top speed the obstacles could reach. Given that, you could use it in any environment, without a map, and you know the UAV is not going to crash into anything.” That guarantee matters because formal safety proofs have long been the missing ingredient in autonomous flight. Trajectory planners typically rely on images and data streamed from a drone’s onboard cameras and sensors to chart a path toward a goal, and most existing approaches fall into one of two camps. Some are designed for unknown but static environments, where obstacles never move. Others handle dynamic obstacles loosely, steering around them heuristically without any formal proof that the robot will not crash. Computing every possible collision scenario in a dynamic world, meanwhile, would take far too long to be useful in real deployments.

“In an unknown dynamic environment, you don’t have many assumptions to rely on,” Kondo explains. “In those types of environments, researchers haven’t yet been able to mathematically guarantee that a trajectory is safe.” The MIT team’s answer was to build rigor into the mathematics from the ground up. Rather than treating safety as a property to be checked after a path is generated, they designed an algorithm whose output is provably safe by construction, covering every possible case, including the awkward edge cases that often break real-world robotic systems. The result is a planner that does not merely hope to avoid obstacles but can demonstrate, mathematically, that it will.

The key to SANDO’s guarantee lies in how it structures the space around the drone. The system begins by mapping out what the researchers call a safety corridor: a series of connected regions of three-dimensional space that the robot can travel through, each of which is guaranteed to contain no obstacles. Safety corridors are not new in themselves, but conventional versions suffer from a fundamental weakness in dynamic settings. A corridor that is clear at one moment can be invaded seconds later by a moving obstacle, silently invalidating the safety proof. SANDO addresses this by making the corridor time-sensitive. Instead of treating obstacles as fixed points, the system explicitly considers their possible future movements when constructing the corridor, so that the guaranteed-clear regions remain valid as the world changes around them.

To make those future predictions possible without knowing exactly where an obstacle will go, SANDO employs a dedicated module that detects, groups, and monitors dynamic obstacles in the environment. For each moving object it identifies, the module uses the obstacle’s maximum velocity to compute how far it could possibly travel within a given timespan. It then encloses the obstacle in a sphere that captures the farthest distance it could reach in any direction over that interval. Even without knowing the obstacle’s actual trajectory, the drone knows the obstacle must remain inside that sphere. SANDO then builds its safety corridor around these spheres, ensuring the planned path can never intersect a region a moving obstacle could occupy. “In the real world, obstacles are going to move, so the safety corridor you create at one point won’t be useful as things move into the corridor,” Kondo says. “But because we consider this time component, we can now guarantee safety into the future.”

Safety alone, however, is not enough if the resulting path is slow or wasteful. To keep flights efficient, SANDO adds a second layer of intelligence: a heat-map-based planner that identifies the “hot” regions of the environment — areas crowded with obstacles — and steers the drone away from them. By biasing the search away from danger zones before the fine-grained trajectory optimization begins, the system charts a more direct and efficient course around hazardous areas rather than threading through them. Once a collision-free safety corridor has been established, SANDO optimizes the trajectory within that corridor to find the fastest possible path to the goal, all while respecting the corridor’s boundaries.

The optimization itself is where much of the engineering effort went. As the robot travels, SANDO continuously adjusts the safety corridor and reformulates the trajectory, ensuring the path remains collision-free all the way to the destination. To make this feasible in real time, the researchers developed a set of mathematical tricks that simplify the optimization problem, allowing the drone’s onboard computer to rapidly recalculate trajectories and react to sudden changes in the environment without relying on offboard processing. The payoff of that rigor, Kondo notes, came in flight testing. “The most difficult part of developing SANDO was the math,” he says. “When you try to guarantee safety, you need to be rigorous and ensure your theory covers every possible case, even edge cases. Once we had that mathematical guarantee, it was very easy to fly the UAVs.”

The experimental results support the theory. In simulation, SANDO reached the robot’s goal faster than several state-of-the-art planning systems while completely avoiding collisions in every environment tested. The team then moved from simulation to the physical world, flying a real UAV equipped with SANDO across twelve test flights. In every one of those flights, the system avoided all dynamic obstacles it encountered, using only the robot’s onboard computer and sensors to replan safe trajectories on the fly. For a field in which simulated guarantees frequently dissolve on contact with real sensors, wind, and latency, that record of zero collisions across both simulated and physical trials is a notable demonstration that the mathematical proof translates into practice.

The potential applications extend well beyond the laboratory. A drone carrying SANDO could fly deep into a wildfire, dodging falling branches and sudden flare-ups while gathering critical information for rescuers on the ground. It could search collapsed buildings for survivors, explore networks of hidden mine tunnels, or deliver packages across a crowded neighborhood where pedestrians, vehicles, and other drones share the airspace. The researchers see several directions for future work: making the system even more computationally efficient, and combining it with machine-learning models that would let users issue instructions to the robot in plain language rather than through technical interfaces. The research was funded in part by the Defense Science and Technology Agency of Singapore. For now, SANDO stands as a rare example of a promise kept in robotics: a planner whose safety is not a hope or a statistic, but a theorem — one that holds even when the world ahead is entirely unknown and everything in it is moving.

Subject of Research: Guaranteed collision-free autonomous trajectory planning for drones in dynamic unknown environments

Article Title: Planning system ensures a robot’s flight path will remain collision-free

Article References: Planning system ensures a robot’s flight path will remain collision-free. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: UAV, trajectory planning, MIT, SANDO, collision avoidance, autonomous navigation, safety corridor, dynamic obstacles, robotics, IEEE Transactions on Robotics, search and rescue, motion planning

The core challenge the researchers set out to solve is twofold. First, aircraft are not free-floating points; they are large, elongated vehicles with real geometry, and they cannot slide sideways or pivot on the spot. Their motion is governed by what roboticists call nonholonomic kinematic constraints, meaning the direction a plane can travel at any instant is tied to the orientation of its nose. Second, when multiple aircraft move at once, their paths and schedules must not intersect in ways that produce collisions, either at the same moment or in sequence as one plane follows another through shared deck space. A planner that ignores either requirement produces trajectories that look elegant on paper but are physically impossible or outright dangerous.

Existing approaches tend to sacrifice one requirement for the other. Classical multi-agent path planning methods, which treat aircraft as abstract tokens moving on a grid or graph, can schedule many agents efficiently but produce paths a real aircraft could never follow, with sharp turns and instantaneous changes of direction. Conversely, kinodynamic planners that respect vehicle dynamics often struggle to coordinate many vehicles at once, because the search space explodes combinatorially as agents are added. The theoretical difficulty is well documented: optimal multi-robot path planning on graphs is computationally intractable in the general case, which has pushed the field toward clever approximations rather than exhaustive search.

HA-SIPP attacks the problem by fusing two established techniques, each compensating for the other’s blind spots. The first ingredient is Hybrid A*, a search algorithm beloved in autonomous driving because it searches over discretized states while propagating motion primitives that obey the vehicle’s actual turning behavior. This ensures every candidate trajectory is something a nosewheel-steering aircraft could genuinely execute. The second ingredient is Safe Interval Path Planning, or SIPP, a framework designed for planning through environments where obstacles appear and disappear over time. SIPP compresses time into safe intervals, windows during which a given region of space is guaranteed to be free of other agents, allowing a single planner to route one vehicle after another through a shared, dynamic map without re-planning everything from scratch.

Embedding Hybrid A* motion primitives inside the SIPP framework is the paper’s central move. The planner searches over a hybrid state space that captures both where an aircraft is and when it will be there, while the safe-interval structure keeps track of which portions of the deck are occupied by already-planned trajectories. When a potential conflict arises between a new trajectory and an existing one, the algorithm does not simply abandon the plan. Instead, it invokes a hybrid validation mechanism that operates on two levels: a fast, conservative screening step followed by a precise geometric certification.

That two-stage safety check is where the discrete-continuous philosophy becomes concrete. In the screening stage, the planner derives conservative collision time windows using the circumradii of the aircraft’s footprint envelopes, essentially circumscribing each aircraft within a circle whose radius guarantees containment of the vehicle’s full outline. If two such circumscribed circles cannot possibly overlap during a given time window, the planner can dismiss the conflict instantly and keep searching. Only when the cheap test flags a possible danger does the algorithm escalate to the second stage: an exact intersection check between the actual polygonal outlines of the two aircraft at the relevant moments. This layered design means the planner pays the computational cost of precise geometry only when it matters, while never certifying a trajectory as safe unless the true shapes have been verified against each other.

The collision model itself is also carefully engineered. Rather than inflating aircraft into oversized bounding circles, which wastes precious deck space and renders tight maneuvers infeasible, the researchers construct a polygon-based collision model that expands the aircraft’s actual geometric contour. The polygon preserves the elongated shape of a carrier aircraft, with its narrow nose and wider wingspan, so that clearance is assessed where the airframe really extends rather than where a crude envelope says it might. The authors note that this preserves shape characteristics while reducing unnecessary conservativeness, a seemingly modest refinement that translates directly into more feasible trajectories in the narrow corridors between deck edge, elevators, catapults and parked aircraft.

To evaluate the approach, the team ran experiments on representative flight-deck layouts with varying fleet sizes, all under fixed computational time budgets, a realistic constraint given that deck operations cannot wait indefinitely for a planner to finish thinking. The results showed that HA-SIPP improved both planning success rates and trajectory feasibility compared with traditional SIPP-based algorithms, and delivered more reliable multi-aircraft conflict avoidance. In practical terms, the hybrid planner was more often able to find complete, executable taxi plans for all aircraft within the allotted time, and the plans it produced were more likely to respect both the kinematic limits of the vehicles and the safety separations demanded by a crowded deck.

The significance of this work extends beyond the carrier deck. The same mathematical structure, kinematically constrained vehicles sharing a confined, time-varying space, describes automated airport ground movement, warehouse fleets of autonomous tugs, and any setting where large vehicles must be marshaled efficiently through tight quarters. The paper situates itself within a rich literature spanning multi-agent path finding with continuous time, prioritized planning for differential-drive robots, and cooperative dispatch frameworks for carrier aircraft, and its contribution is a pragmatic synthesis: keep the scheduling power of safe-interval reasoning, keep the physical realism of hybrid motion primitives, and bind them together with a collision-checking pipeline that is conservative where cheap and exact where necessary.

Challenges remain before such algorithms guide real steel on a real deck. The study is computational, demonstrated on representative layouts rather than at sea, and the authors state that the code is available from the corresponding author on reasonable request, inviting replication and refinement. Real deck operations involve human directors, traction vehicles, weather, and uncertainty in taxi times, factors that the deterministic planner does not fully capture. Yet the direction of travel is clear. As navies pursue autonomous dispatch and as ports and airports automate ground handling, planners like HA-SIPP, which refuse to trade physical feasibility for scheduling convenience, will define the safety envelope within which machines and aircraft share the world’s most congested pavements. The work was partially supported by the National Natural Science Foundation of China under Grant 52171299, and the authors report no conflict of interest.

Subject of Research: Multi-aircraft taxi trajectory planning on aircraft carrier flight decks using hybrid discrete-continuous path planning

Article Title: A Hybrid Discrete-Continuous Strategy for Multi-aircraft Trajectory Planning on Flight Deck

Article References: Fu, W., & Zhu, Q. (2026). A Hybrid Discrete-Continuous Strategy for Multi-aircraft Trajectory Planning on Flight Deck. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01243-2

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01243-2

Keywords: aircraft carrier deck, multi-aircraft taxi planning, trajectory planning, Hybrid A*, Safe Interval Path Planning, collision avoidance, kinematic constraints, nonholonomic motion, multi-agent path finding, autonomous dispatch, motion planning, aerospace engineering

Lung needle biopsies are among the most common and consequential procedures in interventional radiology. When a radiologist spots a suspicious nodule on a chest scan, the definitive diagnosis usually requires extracting a sliver of tissue through the chest wall, guided by computed tomography. The challenge is geometric as much as medical. The needle must reach the lesion while dodging ribs, avoiding the mediastinum and its great vessels, steering clear of pulmonary fissures, and traversing as little healthy lung parenchyma as possible. Every extra millimeter of lung crossed raises the risk of pneumothorax, the collapsed lung that occurs when air leaks through a puncture, as well as bleeding and hemoptysis. Yet despite decades of advances in imaging, the choice of trajectory still varies considerably from one radiologist to the next, a variability that AI and robotics have long been expected to tame.

The new study, led by Taylor Hoffman and Francois H. Cornelis with colleagues including Stephen B. Solomon and Debkumar Sarkar, set out to test whether a general-purpose generative AI model could shoulder part of that planning burden. The team conducted a retrospective, institutional review board-approved analysis of de-identified CT images from 30 consecutive lung biopsies performed by three interventional radiologists, each with at least a decade of experience. The cohort included 17 men and 13 women with a median age of 65 years, and the lesions ranged from 5 to 45 millimeters across, deliberately spanning technically difficult territory: five apical lesions at the top of the lung, seven subpleural nodules hugging the chest wall, three para-mediastinal lesions, and four near the lung base.

The technical setup was elegantly simple. Each axial CT image, acquired on a GE Healthcare scanner and exported in DICOM format, was cropped and magnified but not rescaled. Using the open-source image editor GIMP, the team overlaid a radial grid on the image and converted it to PNG. One author annotated the target lesion with a blue circle and outlined the sensitive structures in red, including the mediastinum and major vessels. The annotated image was then uploaded to ChatGPT-4 with a standardized prompt specifying five key parameters for an optimal entry angle: reaching the target, minimizing the distance traveled through lung tissue, and avoiding bones, fissures, and the mediastinum. The model was asked to read the radial grid and return the angle at which the needle should enter.

Before running the full analysis, the researchers validated the approach by comparing a pre-biopsy planning CT image with a procedural CT image captured during the actual biopsy, acquired at the same anatomical level and showing the needle, skin, and target in place. A separate validation dataset of 30 paired planning and procedural images was then used for data collection. Each case was processed independently in a fresh chat session, with iterations allowed whenever the model failed to incorporate all criteria. The median number of iterations needed to arrive at a complete answer was just two, with an interquartile range of one to three, suggesting the model could converge on a usable plan with minimal back-and-forth.

The results were striking. Independent reviewers, two experienced interventional radiologists who had not performed the original procedures and who assessed the AI-generated paths in blinded fashion, judged 28 of 30 paths, or 93.3 percent, safe and adequate for performing the biopsy. The two rejected paths were deemed theoretically feasible but not representative of best clinical practice: in one case the AI path crossed 20 millimeters of lung tissue, four times longer than the manual route, and in another it unnecessarily punctured lung parenchyma to reach a subpleural nodule, raising concerns about needle stability. More remarkably, the AI paths crossed a median of just 5 millimeters of lung tissue, compared with 23 millimeters for the paths the human radiologists actually chose, a difference that reached statistical significance with a p-value of 0.0063 on a Wilcoxon signed-rank test. Paths with zero lung traversal corresponded exclusively to subpleural nodules, where a direct pleural puncture enters the lesion immediately.

On the question of whether the AI simply rediscovered what the humans had already chosen, the answer was nuanced. The mean entry angle suggested by the AI was 179 degrees, against 174 degrees for the manual procedures, a difference that was not statistically significant. Using a definition of path match requiring entry angles within 10 degrees of each other and similar trajectories through tissue planes, 8 of 30 cases, or 26.7 percent, showed good concordance. The authors note that this match rate is lower than in earlier work, such as a study by Too and colleagues that used a three-dimensional convolutional neural network for lesion detection and path planning and reported 82 percent concordance with an average angular deviation of 2.30 degrees, alongside 93.5 percent sensitivity and 93.2 percent specificity in lesion detection. That earlier system, however, was purpose-built for the task rather than adapted from a general-purpose chatbot, and it defined path match more narrowly, within 5 degrees.

Why should a shorter path matter? The authors argue that reduced traversal through lung tissue could lessen tissue manipulation and lower the risk of pneumothorax and hemoptysis, potentially shortening procedure time, improving patient comfort, reducing anesthesia use, and increasing operating room efficiency. In the 30 procedures studied, no major complications occurred, and minor events were self-limiting: minimal pneumothorax in four cases, 13.3 percent, which resolved spontaneously, and mild hemoptysis in two cases, 6.7 percent. But the authors also flag a clinical caveat. With a coaxial technique, in which an outer introducer needle anchors the pathway for repeated sampling, a very short path through lung tissue might provide insufficient purchase, increasing the risk that the needle dislodges during the procedure. Shorter, in other words, is not automatically better in every scenario.

The study is candid about its limits. With only 30 cases and a retrospective design, generalizability remains unproven. The analysis was performed in two dimensions on single axial slices rather than in three dimensions, a simplification that may underestimate the true spatial complexity of biopsy planning. The prompt did not require the AI to maintain a standard 5-millimeter safety margin from the subclavian and axillary neurovascular bundles, though those structures were kept out of the imaging field because patients were scanned with their arms raised. Iterative follow-up prompts were not scripted, which limits reproducibility, and the study measured technical feasibility only, not clinical outcomes such as diagnostic yield, complication rates, or procedure time. The authors also caution that GPT-4 itself may soon be superseded, arguing that the field should move toward specialized generative models trained specifically on interventional radiology rather than adapting general-purpose systems.

Even so, the trajectory of the technology is hard to ignore. The authors envision generative AI integrated with robotic biopsy platforms to automate trajectory planning, real-time procedural guidance in which the model continuously analyzes intraoperative imaging and suggests adjustments as the patient moves or breathes, pre-procedural virtual rehearsals that let clinicians simulate multiple approaches before committing to one, and eventual three-dimensional reconstruction that would eliminate manual image annotation altogether. Combined with augmented reality, AI-recommended paths could be projected directly into the clinician’s field of view. The same logic extends beyond the lung to tumor ablation, where precise positioning determines whether the entire tumor is covered, and to vascular interventions, where AI could recommend optimal access routes based on patient-specific anatomy. For now, the message of this small feasibility study is measured but provocative: a model built to converse with humans can, with a radial grid and a well-crafted prompt, plan a needle path through the lung that is measurably less invasive than the one a seasoned specialist picked. The needle, for the moment, remains in human hands, but the map may increasingly be drawn by machines.

Subject of Research: Feasibility of using a multimodal large language model to plan entry angles for CT-guided lung needle biopsies

Article Title: Feasibility study on using multimodal large language model for CT-guided lung biopsy trajectory planning

Article References: Hoffman, T., Worthington, M., Sarkar, D., Solomon, S. B., & Cornelis, F. H. (2025). Feasibility study on using multimodal large language model for CT-guided lung biopsy trajectory planning. CVIR Oncology, 1(1), Article 5. https://doi.org/10.1007/s44343-025-00005-3

Image Credits: AI Generated

DOI: 10.1007/s44343-025-00005-3

Keywords: generative AI, ChatGPT-4, lung biopsy, CT-guided biopsy, trajectory planning, interventional radiology, pneumothorax, needle path planning, multimodal large language model, Memorial Sloan Kettering, CVIR Oncology, medical imaging

Hepatocellular carcinoma is the most common form of primary liver cancer and ranks third among all cancers in worldwide mortality. For tumors caught at an early stage, percutaneous thermal ablation, in which a needle-like antenna delivers heat to cook the tumor away, is considered the optimal minimally invasive locoregional therapy. But the technique depends entirely on the operator’s ability to see the tumor. At most institutions, tumors between 3 and 5 centimeters are typically treated with radiation segmentectomy or with transarterial chemoembolization combined with microwave ablation. Both approaches rely on catheters threaded into the arteries feeding the tumor, and both depend on iodinated contrast to map the tumor’s blood supply. When contrast is off the table, those pathways close.

The patient in this report faced exactly that dilemma. His tumor board determined he was ineligible for transplantation, surgical resection, or stereotactic body radiotherapy because of regional lymph node metastases and compromised liver function, classified as Child-Pugh B7 cirrhosis. His stage 4 chronic kidney disease made contrast-enhanced procedures hazardous, and he refused contrast administration altogether. That ruled out radiation segmentectomy, transarterial chemoembolization, and cone-beam CT arteriography, the contrast-enhanced imaging technique that normally guides such procedures in the angiography suite. After extensive discussion, he agreed to a novel alternative: microwave ablation guided by ultrasound-placed fiducial markers and non-contrast cone-beam CT with trajectory planning software.

The technique began with the placement of six radio-opaque fiducial markers, short pushable coils measuring 2 millimeters, delivered through a 15-centimeter, 20-gauge needle under ultrasound guidance. Each coil was positioned at one of the tumor’s anatomic boundaries: cranial, caudal, medial, lateral, anterior, and posterior. Ultrasound and fluoroscopic imaging confirmed that all six markers sat satisfactorily around the lesion. The concept is elegantly simple. Because the tumor itself was poorly visible on non-contrast imaging, the team created an artificial outline of it, framing the cancer with metal beacons that any X-ray-based imaging system could detect. The coils effectively drew the tumor’s borders in a language the treatment equipment could read.

With the tumor framed, the team advanced a microwave antenna under ultrasound guidance until its tip abutted the first fiducial marker. A planning cone-beam CT confirmed the probe’s position, and the images were exported to a standalone desktop running ablation visualization software. That software calculated the maximal yet safe ablation zone for the probe’s position, predicting the volume of tissue that a given energy delivery would destroy, and computed the resulting ablation margin, the buffer of treated tissue extending beyond the tumor edge. The team targeted margins of 0.5 to 1 centimeter, the standard needed to eliminate microscopic tumor extensions that imaging cannot see.

What followed was a carefully choreographed sequence of overlapping ablations. After the first burn, a second probe trajectory was planned using the post-placement cone-beam CT, and the antenna was repositioned with the help of a planned trajectory overlaid on live fluoroscopic guidance. A repeat cone-beam CT verified the new position, the planning software extrapolated the residual tumor still to be treated, and the next trajectory was calculated accordingly. The team worked through the fiducials one by one, ultimately performing eight fiducial-targeted ablations plus two additional burns to confirm adequate margins using the cumulative ablation zones displayed by the software and verified on post-ablation imaging. Because the software mapped every zone and cone-beam CT followed each ablation, no additional immediate postoperative imaging was needed.

The technical footprint of the procedure was modest. Total fluoroscopic time was 18.5 minutes, with a cumulative radiation dose of 1405 milligray. There were no intraoperative or immediate complications. The patient experienced only mild epigastric pain for one week after the procedure and reported no complaints at one-month follow-up. Most strikingly, the initial post-procedural MRI and the most recent 26-month follow-up MRI both demonstrated complete response of the target lesion by modified RECIST criteria, the standardized imaging assessment used to judge treatment success in liver cancer trials. For a tumor larger than 3 centimeters treated without contrast, that durability is the report’s central evidence.

The authors acknowledge meaningful limitations. This is a single-patient retrospective case report, and skeptics could argue that the same outcome might have been achieved with ultrasound or conventional CT alone. But the authors counter that the tumor’s size demanded multiple probe placements and overlapping ablations, which would have been impossible to verify with ultrasound alone because the ablation planning software requires cross-sectional imaging to confirm sufficient coverage. Non-contrast conventional CT, meanwhile, would have limited the precise oblique probe trajectories needed to cover the entire lesion while protecting adjacent bowel. Out-of-plane needle placements are more accurate with the trajectory planning and real-time fluoroscopic guidance that cone-beam CT uniquely provides. The fundamental weakness of non-contrast cone-beam CT is its limited soft tissue resolution, which is precisely the problem the fiducials were designed to solve.

Fiducial markers are not new to interventional radiology. They have previously been used to guide ablation of colorectal metastases, particularly for lesions that disappear on imaging after chemotherapy, leaving nothing visible to target. But little research exists on fiducial placement for prospective ablation guidance in hepatocellular carcinoma, and the authors argue that the effectiveness and reproducibility of this approach make it a worthy addition to the literature. They also sketch a more refined evolution of the technique: a staged workflow in which fiducials are placed first, a preoperative CT or MRI is then acquired, and that study is uploaded and overlaid on the patient in the procedure room. A single fiducial would serve as an internal control for respiration and patient positioning, allowing the ablation zone to be planned from preoperative imaging alone.

The broader significance of the report lies in what it demonstrates about adaptability in image-guided medicine. Predictive ablation software, originally developed to help operators estimate burn volumes in conventional settings, proved capable of anchoring an entire treatment plan around metal markers instead of a contrast-visible tumor. The approach preserves three advantages that are usually lost when contrast is unavailable: accurate margin assessment, safe probe trajectories through difficult anatomy, and confidence that overlapping ablations have covered the target completely. The authors are careful to note that future prospective studies are needed before the technique can be considered standard practice, and a single case cannot establish safety across diverse patients and tumor geometries. Still, for the growing population of patients with liver cancer and kidneys that cannot tolerate contrast, the report offers a concrete, technically grounded path forward, one in which six small coils and clever software stand in for the contrast dye that medicine long took for granted.

Subject of Research: Fiducial marker-assisted, contrast-free microwave ablation of hepatocellular carcinoma using cone-beam CT guidance and ablation predictive software

Article Title: Fiducial marker-assisted microwave ablation of hepatocellular carcinoma: achieving margins without contrast

Article References: Daniel, B., & Kouri, B. (2025). Fiducial marker-assisted microwave ablation of hepatocellular carcinoma: achieving margins without contrast. CVIR Oncology, 1(1), Article 16. https://doi.org/10.1007/s44343-025-00013-3

Image Credits: AI Generated

DOI: 10.1007/s44343-025-00013-3

Keywords: hepatocellular carcinoma, microwave ablation, fiducial markers, cone-beam CT, interventional radiology, ablation predictive software, contrast contraindication, liver cancer, trajectory planning, cirrhosis, mRECIST, case report

The difficulty begins with the aircraft themselves. Unlike multirotor drones, which can hover, pivot in place, and creep toward a target, fixed-wing UAVs must keep moving forward to stay aloft. They cannot stop, and their turning radii are bounded by aerodynamic reality: bank too sharply and the aircraft stalls or exceeds structural limits. That means every approach to a rescue zone is a commitment. A drone that lines up from the wrong heading may need a wide, time-consuming arc to correct itself, and in a rescue scenario, those seconds compound. Add a moving target, several aircraft with different capabilities and payloads, and the requirement that resources arrive from multiple directions simultaneously to maximize the chance of reaching survivors spread across the water, and the problem becomes a combinatorial nightmare.

Formally, the researchers frame each rescue task as a specific approach direction, or AD, for an individual FW-UAV to deploy its resources. Assigning those directions across a heterogeneous fleet is a high-dimensional optimization problem with a multimodal landscape, meaning the search space is studded with many locally good solutions that can trap conventional algorithms. Classical genetic algorithms and similar evolutionary methods often stagnate in these traps, converging on a decent but far-from-optimal assignment plan while better configurations sit unexplored elsewhere in the solution space. For a time-critical operation like maritime rescue, stagnation is not just an inefficiency; it is a direct cost in human survival probability.

To break out of those traps, the team built an adaptive quantum-inspired genetic algorithm, or AQIGA, as the upper layer of their framework. Quantum-inspired evolutionary algorithms, first popularized in the early 2000s, borrow a mathematical vocabulary from quantum mechanics without using any actual quantum hardware. Instead of encoding candidate solutions as fixed strings, they encode each element as a pair of probability amplitudes, conceptually akin to a qubit existing in a superposition of the zero and one states. A population of these probabilistic representations is measured to produce concrete solutions, evaluated, and then updated by rotation gates that shift the probabilities toward promising regions. The effect is that a single individual can implicitly represent multiple candidate configurations at once, giving the search a richer exploration dynamic than a standard genetic algorithm of the same population size.

The adaptive part of AQIGA is where the new work earns its name. The algorithm continuously monitors two statistics of its population: entropy, a measure of how varied the solutions are, and diversity, a measure of how spread out the population is across the search space. A dynamic feedback loop based on these metrics regulates how aggressively the algorithm explores versus exploits. When the population collapses toward uniformity, a signature of premature convergence, the feedback mechanism pushes the probability updates to broaden the search again; when a healthy spread is present, it tightens focus around the best candidates. This self-regulation is designed specifically to prevent the local optima stagnation that plagues cooperative task assignment problems with many near-equivalent solutions.

Deciding who goes where is only half the battle. The lower layer of the framework is a vector field guidance model, VFGM, that converts each abstract assignment into a flyable, collision-free trajectory. Vector field guidance has a respected lineage in autonomous flight: Lyapunov vector fields, described in seminal work in the Journal of Guidance, Control, and Dynamics in 2008, have long been used for standoff tracking of moving targets, where an aircraft follows a smooth vector field that spirals it onto a desired orbit. The VFGM here extends that idea to a cooperative arrival problem. It steers each FW-UAV along its assigned approach direction while respecting the aircraft’s continuous-forward-motion requirement and limited turning radius, deconflicting the trajectories so that multiple drones converge on the rescue area from different headings at the same time without crossing paths dangerously close to one another or to the distressed vessel.

The simultaneous multidirectional arrival is the crux of the rescue logic. Survivors in the water scatter, and the drifting life raft or the distressed vehicle may have people clinging to different sides. Resources dropped from a single heading cover only a limited arc, and in rough seas, timing matters as much as placement. By synchronizing arrival from multiple directions, the fleet blankets the area rather than lining it. The decentralized nature of the VFGM also matters operationally: rather than a single ground station computing every path for every aircraft, each UAV executes its own guidance in real time, which makes the system more robust to communication dropouts, a serious concern over open ocean where satellite links are the only option and bandwidth is precious.

To find out whether the design actually works, the researchers tested it in simulated maritime rescue scenarios featuring moving target vehicles, and benchmarked the AQIGA against conventional quantum-inspired algorithms under the same conditions. The results were consistent and, in places, striking. On average, the full AQIGA-VFGM framework improved solution fitness by 13.23 percent, meaning the task assignments it produced were substantially better by the combined objectives of the mission. It cut convergence time by 37.36 percent, which in a rescue context translates directly into faster planning when every minute of drift separates survivors from rescuers. And it improved what the authors call excess resource fairness by 10.88 percent, a metric reflecting how evenly the burden of resource deployment is distributed across the fleet, preventing scenarios where one or two aircraft are overloaded while others cruise underutilized.

Those three numbers tell a coherent story. Fitness improvement means the plans are better; convergence improvement means they arrive sooner; fairness improvement means the fleet’s collective capacity is actually used. The authors argue that together these gains make the framework promising and practically feasible for time-critical multi-UAV maritime rescue operations. It is worth noting, as the study itself does, that the validation so far is in simulation. Moving from simulated moving-target scenarios to real fixed-wing aircraft over real seas will demand field trials that stress communication links, sensor error, weather, and the safety case for autonomous drops near people in the water. The lineage of the underlying guidance methods, however, is well established in flight trials elsewhere, which gives the simulation results a credible foundation.

The broader significance stretches beyond maritime rescue. The core architecture, a quantum-inspired adaptive optimizer handing assignments to a decentralized vector field guidance layer, is modular by design. The same pattern could apply to coordinated wildfire suppression drops, distributed package delivery, disaster inspection with mixed drone fleets, or any mission where heterogeneous fixed-wing vehicles must divide spatially structured tasks and execute them in concert against moving targets. The authors’ own bibliography traces a decade of progress in cooperative task assignment, from multi-type genetic algorithms for heterogeneous UAV teams to consensus-based dynamic assignment under uncertainty, and their contribution slots neatly into that trajectory by attacking the two constraints that fixed-wing platforms impose and that rotorcraft-centric research has often been able to ignore. As drone fleets grow larger and missions grow more complex, frameworks like this one, which marry a principled optimization scheme with flight-physics-aware execution, may become the standard way that swarms of ordinary, aerodynamically constrained aircraft manage to behave like an extraordinary, coordinated team.

Subject of Research: Cooperative task assignment and execution for heterogeneous fixed-wing UAV swarms in maritime rescue

Article Title: Cooperative Task Assignment and Execution for Heterogeneous Fixed-Wing UAVs in Maritime Rescue

Article References: Baig, M. I., Zhen, Z., & Javaid, U. (2026). Cooperative Task Assignment and Execution for Heterogeneous Fixed-Wing UAVs in Maritime Rescue. International Journal of Aeronautical and Space Sciences. https://doi.org/10.1007/s42405-026-01273-w

Image Credits: AI Generated

DOI: 10.1007/s42405-026-01273-w

Keywords: fixed-wing UAVs, maritime rescue, quantum-inspired genetic algorithm, vector field guidance, task assignment, multi-UAV coordination, heterogeneous drone fleet, simultaneous arrival, optimization, autonomous flight, search and rescue, trajectory planning

Liver biopsy is a cornerstone of cancer diagnosis, but it is far from trivial. The liver is a crowded, vascular organ hemmed in by ribs, lungs, bowel, and major blood vessels, and the needle must reach a target lesion while avoiding all of these hazards. Today, that planning rests almost entirely on the experience and judgment of individual interventional radiologists, which introduces meaningful variability between operators. Standardized AI-assisted planning has been proposed as a way to reduce that variability, particularly at centers with less experienced staff, but the field remains preliminary. The Memorial Sloan Kettering team set out to test the most basic building block of such a system: could a generative AI model, given nothing but a single axial CT slice and a standardized prompt, propose an entry angle that experienced reviewers would rate as potentially safe?

The study design was deliberately simple. The researchers retrospectively collected de-identified CT images from 30 consecutive CT-guided liver biopsies performed between February and June 2025 by sixteen board-certified interventional radiologists with a median of eight years of experience. The patient group included 20 women and 10 men with a median age of 56, and their lesions spanned liver segments 1 through 8 under the Couinaud classification. Crucially, the AI never touched a patient. It worked only on the initial planning CT image, after the fact, and only in two dimensions, while the actual procedures had already been completed safely by human hands.

The preprocessing workflow was strikingly low-tech for a study of cutting-edge AI. Each CT image was cropped to remove patient identifiers, magnified without rescaling, and overlaid with a radial grid created in GIMP, the free open-source image editor. The grid consisted of 36 radial lines at 10-degree intervals with concentric circles spaced 10 millimeters apart, calibrated to known anatomical landmarks. The annotated images were saved as PNG files and fed to GPT-4.1 through a standardized prompt. The model operated entirely on its baseline training, with no fine-tuning on medical data, and a fresh chat was started for every case to prevent contamination between cases.

The researchers tested two conditions that differed in how much hand-holding the AI received. In the first condition, two study authors manually annotated each image with a blue circle marking the lesion and a red line delineating sensitive structures to avoid, essentially handing the model a pre-drawn map of the danger zones. In the second, more demanding condition, only the blue circle was drawn, forcing the AI to identify critical structures on its own. For each case, the model was allowed up to four iterations: if its proposed trajectory failed safety criteria, it was re-prompted within that limit. Two independent reviewers then rated every AI-generated trajectory using consensus methodology, judging safety, avoidance of organs, vessels, and bone, whether the path crossed the midline or excessive tissue, and whether the approach would be comfortable and clinically reasonable.

The headline numbers were encouraging, with important caveats. In the first condition, the AI needed a median of two iterations and produced trajectories rated potentially safe in 27 of 30 cases, or 90 percent. In the second condition, without the red-line annotations, it needed a median of just one iteration and scored 28 of 30, or 93 percent. The difference between conditions was not statistically significant, and the model’s proposed entry angles differed significantly from the angles actually used by clinicians, with mean differences reaching statistical significance in both conditions. Only 23 percent of cases in condition one and 33 percent in condition two showed strong concordance, defined as entry angles within 10 degrees with similar liver traversal patterns, meaning the AI was often safe but not necessarily matching human choices.

The failures are as instructive as the successes. In three cases in the first condition, the AI proposed paths crossing bone, the midline, or the marked red line, even after four attempts. One of these involved a patient with ascites, fluid in the abdomen that complicated the anatomy; another involved a patient whose prior surgery had removed the left hepatic lobe, creating atypical anatomy; and a third targeted a high-lying segment 2 lesion where the model routed the needle through costal cartilage. Cartilage proved a recurring blind spot: in the second condition, both unsafe trajectories involved bone or cartilage, suggesting the model struggles to recognize structures with intermediate radiodensity. The authors note that excluding these inadequate cases from distance calculations introduces selection bias that may flatter the AI’s performance.

On the metric of liver tissue traversed, the AI actually looked competitive. In condition one, its paths crossed a median of 31.6 millimeters of liver versus 48.25 millimeters for the manual paths, and in condition two, 39.35 versus 48.25 millimeters, though neither difference reached statistical significance. Interestingly, the AI’s longer median distances in the unannotated condition hint at a conservative streak: without predefined avoidance zones, it may favor longer but safer routes. A power analysis indicated the 30-case sample could only detect large effect sizes, so smaller but clinically meaningful differences would require larger cohorts to detect.

The authors are refreshingly blunt about what these results do not mean. This was a proof of concept, nothing more. The workflow does not address a real clinical need: interventional radiologists can identify safe trajectories in seconds using native console capabilities with full three-dimensional visualization, whereas the AI pipeline required manual image export, GIMP annotation, and chatbot prompting, all while discarding the three-dimensional spatial information that real planning depends on. A human planner scrolls through multiple slices, uses multiplanar reconstruction, and mentally constructs 3D anatomy, capabilities entirely absent from a static 2D analysis. The AI also required continuous human supervision, with iterative re-prompting in most cases, which undermines any claim of autonomy. And because the AI trajectories were never actually executed, they remain theoretical, while the manual paths were proven safe by successful, complication-free procedures, a comparison that cannot establish equivalence.

Looking forward, the researchers see a path, if a long one. Future systems would need automated lesion identification, since current generative models lack the segmentation capabilities required for precise localization, and integration with dedicated segmentation architectures such as U-Net could enable fully autonomous planning from raw CT images. Three-dimensional volumetric analysis, quantitative vascular proximity measurements, phantom studies, and prospective multicenter trials would all be prerequisites before any clinical consideration. The authors also point toward applications in tumor ablation and margin assessment, and note that many liver biopsies worldwide are performed under ultrasound guidance, where real-time, operator-dependent imaging poses an even harder challenge for AI. For now, the message is measured: generative AI can look at a liver CT slice and propose a plausible needle angle most of the time, but the 7 to 10 percent of cases where it fails, sometimes by routing a needle through bone or cartilage, are exactly the cases where a patient would be harmed. In medicine, that margin is everything.

Subject of Research: Feasibility of generative AI for planning CT-guided liver biopsy needle trajectories

Article Title: Feasibility of generative AI for CT-guided liver biopsy trajectory planning: a pilot study

Article References: Feasibility of generative AI for CT-guided liver biopsy trajectory planning: a pilot study. (n.d.). https://doi.org/10.1007/s44343-026-00041-7

Image Credits: AI Generated

DOI: 10.1007/s44343-026-00041-7

Keywords: generative AI, GPT-4.1, liver biopsy, CT guidance, interventional radiology, trajectory planning, pilot study, medical imaging, needle biopsy, proof of concept, Memorial Sloan Kettering, artificial intelligence