In the aftermath of a devastating earthquake, unmanned aerial vehicles (UAVs) can fly through a collapsed building to map the scene and give rescuers the information they need to quickly reach survivors.
However, this remains an extremely challenging problem for an autonomous robot, which would need to quickly adjust its trajectory to avoid sudden obstacles while staying on course.
Researchers from MIT and the University of Pennsylvania have developed a modern trajectory planning system that addresses both challenges simultaneously. Their technique enables the UAV to react to obstacles within milliseconds while maintaining a velvety flight path, which minimizes travel time.
Their system uses a modern mathematical formulation that ensures the robot will safely reach its destination along a feasible path, while requiring less computation than other techniques. In this way, it generates smoother trajectories faster than state-of-the-art methods.
The trajectory planner is also powerful enough to enable real-time flight using only the robot’s onboard computer and sensors.
The open-source system, called MIGHTY, does not require proprietary software packages that can cost hundreds of thousands of dollars. It can be more easily implemented in a wider range of real-world settings.
In addition to search and rescue, MIGHTY can be used for applications such as last-mile deliveries in urban spaces, where UAVs must avoid buildings, wires and people, or for industrial inspections of convoluted structures such as wind turbines.
“MIGHTY achieves comparable or better performance by using only open source tools, which means that any researcher, student, or company – anywhere in the world – can use it freely. By removing this cost barrier, MIGHTY helps democratize high-performance trajectory planning and opens the door to a much broader community that can leverage this work,” says Kota Kondo, an aeronautics and astronautics graduate student and lead author of the paper on this planning tool trajectory.
In the article, Kondo is joined by Yuwei Wu, a graduate of the University of Pennsylvania; Vijay Kumar, professor at UPenn; and senior author Jonathan P. How, Ford Professor of Aeronautics and Astronautics and principal investigator at the Laboratory for Information and Decision Systems (LIDS) and the Aerospace Control Laboratory (ACL) at MIT. Tests appears in .
Overcoming compromises
When Kondo was a child, the Fukushima Daiichi nuclear accident occurred after the Great East Japan Earthquake. After school was canceled, Kondo stayed home and watched the news every day while workers surveyed and secured the reactor site. Some workers still had to enter hazardous areas to contain the damage and assess the situation, exposing them to high doses of radioactive materials.
“I was passionate about creating autonomous robots that can enter dynamic and dangerous situations and then come back and report back to humans who stay out of danger,” Kondo says.
This task requires a good trajectory planning tool, which is software that determines the path the robot should follow to get from point A to point B safely.
However, many existing systems impose trade-offs that limit performance.
Although some commercial systems can quickly generate velvety trajectories, they can cost hundreds of thousands of dollars. Open source alternatives are often weaker compared to commercial solutions or challenging to exploit.
With MIGHTY, Kondo and his colleagues have developed an open-source system that generates high-quality, velvety trajectories by responding to obstacles in real time and that runs speedy enough to fly using only on-board components.
To do this, they overcame a key challenge that limits many open source systems.
These methods usually first estimate how long it will take the robot to get from point A to point B. Based on the estimated travel time, the planner finds the best way to reach the destination.
Although using a fixed travel time allows the planner to quickly generate a trajectory, it has drawbacks. First, if the UAV needs to move very far away from obstacles to avoid obstacles, it may have to boost its speed to meet the set travel time budget. This makes it challenging to avoid sudden threats.
A STRONG method
Instead, MIGHTY uses a mathematical technique called the Hermite spline, which optimizes travel time and flight path in one step, creating a velvety trajectory that can be precisely controlled.
“Combined optimization of the spatial and temporal components allows us to get better results, but now the optimization is so large that it is more difficult to solve in the possible time,” Kondo says.
The researchers used a clever technique to reduce this computational overhead.
Instead of generating a trajectory from scratch every time, MIGHTY pre-guesses the trajectory. It then refines the trajectory through iterative optimization using the scene map generated by the UAV’s lidar sensors.
“We can reasonably guess what the trajectory should be, which is much faster than generating the whole thing from nothing,” Kondo says.
This allows MIGHTY to react in real time to unknown obstacles, while maintaining a velvety trajectory and minimizing travel time. The system uses on-board UAV components, which is crucial in applications where the robot can move far from the base station.
In simulated experiments, MIGHTY required only about 90 percent of the computational time required by state-of-the-art methods and safely arrived at the target about 15 percent faster than those approaches.
When they tested the system on real robots, it reached speeds of 6.7 meters per second, avoiding every obstacle that came in its path.
“With MIGHTY, everything is integrated into one whole. There is no need to communicate with any other software to find a solution. This allows us to work even faster than some commercial solutions,” says Kondo.
In the future, researchers want to improve MIGHTY so it can be used to control multiple robots at once and conduct more flight experiments in challenging environments. They hope to continue to improve the open source system based on user feedback.
“MIGHTY makes an crucial contribution to agile robot navigation by rethinking the trajectory representation itself. Hermite splines have already been used successfully in visual simultaneous localization and mapping, and it is nice to see that their advantages are now being exploited for trajectory planning in mobile robots. By enabling joint optimization of path geometry, time, velocity and acceleration while retaining local control over the trajectory, MIGHTY gives robots greater freedom to calculate speedy, dynamically feasible movements in cluttered environments,” says Davide Scaramuzza, professor and director of the Robotics and Perception Group at the University of Zurich, who was not involved in this research.
This research was funded in part by the U.S. Army Research Laboratory and the Defense Science and Technology Agency in Singapore.