MIT researchers have developed an algorithm that enables drones to avoid collisions while working together in the same airspace. The system, known as Robust MADER, is an improved version of the Multiagent Trajectory-Planner, or MADER, which was presented by MIT researchers in 2020.
MADER enables a group of drones to create optimal, collision-free trajectories by broadcasting their intended paths and considering each other’s plans when optimising their own. However, the researchers found that when the system was tested on real drones, communication delays between agents caused some drones to select paths that could result in collisions.
To address this issue, the researchers developed Robust MADER, which generates collision-free trajectories even when communications between agents are delayed. The algorithm includes a delay-check step, during which a drone waits a specific amount of time before committing to a new trajectory. If it receives additional trajectory information from fellow drones during the delay period, it may abandon its new trajectory and start the optimisation process over again.
Kota Kondo, an aeronautics and astronautics graduate student, said that the new system achieved a 100% success rate in generating collision-free trajectories when tested in simulations and real-world experiments with drones. While the drones’ travel time was slightly slower than it would be with other approaches, safety was the top priority. Kondo added that “if you collide with something, no matter how fast you go, it doesn’t really matter because you won’t reach your destination.”
The Multiagent Trajectory-Planner is asynchronous, decentralised and multi-agent, meaning each drone creates its own trajectory, and all agents must agree on each new trajectory, but they don’t need to agree at the same time. This makes MADER more scalable than other approaches, which would struggle to get thousands of drones to agree on a trajectory simultaneously.
Each drone optimises a new trajectory using an algorithm that incorporates trajectories received from other agents. By continually optimising and broadcasting their new trajectories, the drones avoid collisions. However, in real-world environments, signals are often delayed by interference from other devices or environmental factors such as stormy weather. Robust MADER prevents collisions because each agent has two trajectories available. It keeps one trajectory that it knows is safe, and while following that trajectory, the drone optimises a new trajectory but doesn’t commit to the new path until it completes a delay-check step.
The length of the delay-check period depends on the distance between agents and environmental factors that could hamper communications. If the agents are many miles apart, for instance, then the delay-check period would need to be longer.
The researchers tested Robust MADER in a multiagent flight environment, using six drones and two aerial obstacles. While the original version of MADER would have caused seven collisions, Robust MADER did not cause a single crash in any of the hardware experiments.
Kondo said that seeing the system working in practice was rewarding, adding that “until you actually fly the hardware, you don’t know what might cause a problem. Because we know that there is a difference between simulations and hardware, we made the algorithm robust, so it worked in the actual drones.”
The research will be presented at the International Conference on Robots and Automation.