Radar Odometry Subject to High Tilt Dynamics of Subarctic Environments

A team of researchers has published a paper introducing a novel radar-inertial odometry method designed to solve a critical robotics problem: navigation in highly dynamic, uneven terrain. The work, led by Matěj Boxan, William Larivée-Hardy, and François Pomerleau, specifically targets subarctic environments where standard methods fail. These methods typically assume flat ground, but the team's data shows real-world conditions with consecutive scan tilts reaching 13° in pitch and 4° in roll, with absolute vehicle pitch hitting a dramatic 30°.

The researchers' new algorithm introduces two key innovations to handle these extremes. First, it employs a 'tilt-proximity submap search' to better match radar scans as the vehicle's orientation changes drastically. Second, it implements a hard threshold for vertical displacement between scan points and the estimated axis of rotation, filtering out erroneous data from steep slopes. The team benchmarked their method against three existing state-of-the-art radar odometry techniques on an urban baseline and a demanding 2-kilometer 'dynamic trajectory.' Their results show state-of-the-art performance on the urban test and a 0.3% accuracy improvement on the long, challenging run.

Finally, the paper provides a detailed performance analysis on a complex sequence involving high lateral slip and a steep ditch traversal, scenarios that cripple conventional approaches. This work directly addresses a limitation for autonomous systems in mining, polar research, and off-road exploration, proving that robust navigation is possible even when the ground is anything but flat.