Drone and Autonomous Vehicle Tracking at M-Air
M-Air is a 10,000 square-foot, four-story, netted scientific facility that enables the study of autonomy and collaborative robotics in the complete range of Michigan weather, from below 0°F /-18°C winters to high humidity summers with all the snow, rain, hail, and sleet that comes throughout the year.
The Qualisys 7+ cameras allow the group to study outdoors and in the extreme weather condition. Currently, the space holds 34 Qualisys 7+ cameras track and explore autonomous flight/vehicular activity.
A Space for Robotic Discovery
M-Air is located next door to the new Ford Robotics Building, which will be completed in 2020 and offers a home to roboticists from across the University to collaborate and contribute to one another’s work, enhancing and accelerating the robotic solutions that can serve society. Also in the neighborhood is Mcity, the autonomous driving testing ground, wind tunnels for aerodynamic testing, and water tanks for on and underwater testing.
M-Air is located next door to the new Ford Robotics Building, which will be completed in 2020 and offers a home to roboticists from across the University to collaborate and contribute to one another’s work, enhancing and accelerating the robotic solutions that can serve society. Also in the neighborhood is Mcity, the autonomous driving testing ground, wind tunnels for aerodynamic testing, and water tanks for on and underwater testing.
The outdoor location allows the group to test algorithms in more realistic environments than can be simulated indoors. This includes testing quadrotors in fairly windy conditions and testing ground robots on outdoor surfaces, among others.
Ground Truth Data
One particular experiment, submitted by Robotics PhD candidate Ross Hartley, was performed to analyze the accuracy of a state estimator (contact-aided invariant extended Kalman filter) running on the robot. Motion capture from M-Air served as ground truth.
While the subject was walking, the system was estimating the 3D orientation, position, and velocity of the robot using IMU, joint encoder, and contact measurements, presented with several plots attached comparing the state estimates vs. the ground truth from M-Air’s motion capture system
Research at M-Air involves resilient algorithms which preserve the safety and security of multi-agent systems. They study algorithms which allow robots to achieve certain goals even if some of the robots are hacked, faulty, or broken. There are various angles this topic can be approached from, in a control theoretic perspective.
There are several platforms used in the research described by Aerospace PhD Candidate, James Usevitch, including ground robots (seven total) and quadrotors (a lot!). The lab space itself is too small to use very many agents simultaneously, so M-Air provides more space for performing complex tasks with multi-agent teams.
Another ongoing project submitted by Robotics PhD student, Matt Romano, works with cooperative payload transport using an multirotors for manufacturing — specifically, an octocopter with a nailgun to shingle roofs.
For us, the key benefit of M-Air is the large outdoor space the robots can operate in. We tested several indoor motion capture setups before this, but the coverage was too small for the robot to easily walk around in. The huge space in M-Air allowed us to test longer experiments without having to worry about accidentally leaving the area, which is all covered by the motion capture system.
Ross Hartley
Meet Cassie Blue
Cassie Blue is the bipedal robot that helps U-M roboticists take independent robotic walking to a whole new level. Cassie has control over two more joints in each leg – motors for hip rotation and at the ankle. Not only do these give Cassie the potential to be better at the independent walking pioneered by its predecessor, MARLO, but it opens a host of new possibilities.
Want to know more?
Read more about M-Air by following the links below.
M-Air at the University of Michigan
M-Air: Outdoor lab for flying drones and autonomous aerial vehicles
