Motion Planning for Safe Landing of a Human-Piloted Parafoil

Most skydiving accidents occur during the parafoil-piloting and landing stages and result from human lapses in judgment while piloting the parafoil. Training of novice pilots is protracted due to the lack of functional and easily accessible training simulators. Moreover, work on parafoil trajectory planning suitable for aiding human training remains limited. To bridge this gap, we study the problem of computing safe trajectories for human-piloted parafoil flight and examine how such trajectories fare against human-generated solutions. For the algorithmic part, we adapt the sampling-based motion planner Stable Sparse RRT (SST) by Li et al., to cope with the problem constraints while minimizing the bank angle (control effort) as a proxy for safety. We then compare the computer-generated solutions with data from human-generated parafoil flight, where the algorithm offers a relative cost improvement of 20%-80% over the performance of the human pilot. We observe that human pilots tend to, first, close the horizontal distance to the landing area, and then address the vertical gap by spiraling down to the suitable altitude for starting a landing maneuver. The algorithm considered here makes smoother and more gradual descents, arriving at the landing area at the precise altitude necessary for the final approach while maintaining safety constraints. Overall, the study demonstrates the potential of computer-generated guidelines, rather than traditional rules of thumb, which can be integrated into future simulators to train pilots for safer and more cost-effective flights.

💡 Research Summary

The paper addresses the high incidence of sky‑diving accidents that occur during the parafoil piloting and landing phases, which are largely attributed to human judgment errors. Because novice sky‑divers lack access to functional simulators, training relies on informal “rules of thumb,” resulting in prolonged skill acquisition and increased risk. The authors therefore propose a method to automatically generate safe landing trajectories for human‑piloted parafoils and evaluate these trajectories against real flight data.

A simplified four‑state dynamic model is adopted, where the state vector consists of horizontal position (x, y), altitude (h), and heading (ψ). The only control input is the bank angle ϕ, which directly influences the glide angle and turn rate. Safety is encoded in a cost function J = ∫ ϕ² dt, i.e., the time‑integrated square of the bank angle. Minimizing this cost reduces both the number of turns and the duration of each turn, thereby limiting vertical speed spikes that are dangerous at low altitude.

To solve the resulting kinodynamic planning problem, the authors adapt the Stable Sparse RRT (SST) algorithm, an anytime, asymptotically optimal sampling‑based planner. Two key modifications are introduced: (1) the nearest‑neighbor search is performed only in the three‑dimensional physical space (x, y, h), deliberately ignoring heading ψ so that the planner focuses on spatial proximity rather than orientation; (2) during the final approach phase (the last ≈ 160 m of altitude) a deterministic proportional controller replaces random bank‑angle sampling. This controller computes ϕ = 0.5 (ψ − λ_w), where λ_w is the wind direction, forcing the parafoil to fly straight against the wind and preventing sharp low‑altitude turns.

Implementation uses the Open Motion Planning Library (OMPL) with Python bindings, and trajectory propagation is performed with the DOPRI integrator from the diffrax library. Simulations run on a standard laptop (Intel Core Ultra 7, 4.8 GHz, 32 GB RAM) without GPU acceleration. A representative scenario places the parafoil at (−1500 m, 500 m, 1000 m) with speed 19.2 m/s and glide ratio 3, aiming for a 400 m × 400 m landing zone centered at the origin and 5 m above ground. The planner finds a collision‑free path within 30 seconds, and the resulting cost is 20 %–80 % lower than the average cost recorded from human pilots performing the same task.

Analysis of solution quality versus planning time shows the expected SST behavior: as more iterations are allowed, the cost monotonically improves, converging toward the optimum. The proportional controller in the final approach successfully limits altitude error to less than 0.5 m and eliminates abrupt vertical velocity increases, which are the primary cause of hard landings.

The authors acknowledge several limitations. Wind speed and direction are assumed constant, ignoring real‑time turbulence and gusts. The model presumes instantaneous bank‑angle execution, whereas actual pilots experience latency and physiological constraints. Moreover, the cost function focuses solely on control effort and does not directly penalize flight time or fuel consumption. Future work is outlined to incorporate stochastic wind models, real‑time augmented‑reality feedback, multi‑pilot cooperative planning, and reinforcement‑learning‑based adaptive controllers.

In conclusion, the study demonstrates that a suitably adapted SST planner can generate smoother, safer landing trajectories for human‑piloted parafoils, achieving substantial reductions in control effort compared with human‑generated flights. These computer‑generated trajectories can serve as quantitative training guidelines, offering a promising component for next‑generation sky‑diving simulators and pilot education programs.