Researchers from MIT and the University of Pennsylvania have developed a new trajectory planning system for unmanned aerial vehicles (UAVs) that outperforms commercial alternatives and costs nothing to deploy. Named MIGHTY, the system is fully open-source and requires no proprietary software packages. The results were published on June 17, 2026 in IEEE Robotics and Automation Letters.
Key takeaways
- MIGHTY uses Hermite splines to jointly optimize path geometry and travel time — 15% faster goal arrival than state-of-the-art methods
- The system runs entirely on the drone's onboard computer with no external server required
- In real-world flights, the drone reached 6.7 m/s while avoiding every obstacle encountered
- MIGHTY requires roughly 90% of the computation time used by competing state-of-the-art methods
- Fully open-source — eliminates the need for commercial solvers costing hundreds of thousands of dollars
The bottleneck: separating path from time
Most existing UAV trajectory planners treat travel time as a fixed input — they estimate flight duration first, then compute the path. This two-step approach forces the robot to stick to a predetermined speed budget even when new obstacles appear, sometimes requiring it to accelerate through a hazard zone to stay on schedule.
MIGHTY eliminates this split by optimizing the spatial path and the temporal profile in a single step. The mathematical backbone is a Hermite spline representation, which allows the planner to simultaneously control path shape, velocity, and acceleration while maintaining smooth, physically executable trajectories.
The computational challenge is real: joint optimization of space and time is harder to solve than either alone. MIGHTY addresses this through an iterative refinement strategy — rather than generating a trajectory from scratch on each sensor update, the planner starts from an initial guess and refines it using a lidar-based map built onboard. This keeps the system fast enough for real-time flight on commodity hardware.
Results: faster and cheaper than commercial alternatives
In simulation, MIGHTY reached its destination 15% faster than comparable methods while requiring only 90% of their computation time. In physical flight tests, the drone hit 6.7 m/s without a single collision, including when unexpected obstacles appeared mid-flight.
The cost angle is equally significant. Commercial optimization solvers used by similar systems carry licensing fees ranging from tens to hundreds of thousands of dollars. MIGHTY is built entirely from open-source libraries. Any researcher, student, or company anywhere in the world can download and deploy it without licensing costs — a point the lead author Kota Kondo explicitly emphasized as a design priority.
Context and limitations
Real-time trajectory planning remains one of the harder problems in mobile robotics. Systems must react to environmental changes in milliseconds while keeping motion smooth enough for physical actuators to follow. Prior approaches from groups such as Berkeley AI Research Lab and MIT's own Aerospace Controls Laboratory achieved strong results in simulation but required external solvers or powerful offboard servers.
MIGHTY does not solve every problem. The current system handles single-robot planning only — multi-robot or swarm coordination is out of scope for this version. Physical tests were conducted in a controlled environment, and performance in chaotic real-world conditions remains unvalidated. That said, the combination of onboard computation, open-source tooling, and competitive benchmark performance positions MIGHTY as a credible candidate for deployment in contexts where commercial solvers were previously prohibitive.
Why this matters
Trajectory planning is the backbone of mobile robot autonomy. Every rescue drone, inspection robot, and last-mile delivery UAV needs a system that recomputes an optimal path in fractions of a second as sensor data changes. The commercial solver barrier has historically excluded smaller labs, university teams, and organizations in lower-income regions from deploying competitive autonomous systems.
MIGHTY's novelty is not a physics breakthrough — it is the engineering combination of known mathematical tools (Hermite splines, iterative optimization) in a way that eliminates performance bottlenecks and removes the need for expensive licenses. If real-world results hold in harder conditions, the system could become a standard building block for next-generation rescue and delivery robots.
What's next
- The MIT Aerospace Controls Laboratory team plans to extend MIGHTY to multi-robot coordination — the next iteration of the research agenda
- Real-world validation in disaster-scenario environments (post-earthquake rubble, dynamic obstacles) is the prerequisite for commercial rescue deployment
- Code and documentation are available alongside the IEEE Xplore publication in Robotics and Automation Letters
Sources
- Robohub — New research enables a robot to chart a better course
- IEEE Robotics and Automation Letters — MIGHTY: Minimum-Time Trajectory Planner (Kondo et al., 2026)
