Online identification of skidding modes with interactive multiple model estimation

TL;DR

This work proposes an interactive multiple model (IMM) based filtering framework to probabilistically identify predefined robot operation modes that could arise due to traversal in different terrains or loss of wheel traction.

Abstract

Skid-steered wheel mobile robots (SSWMRs) operate in a variety of outdoor environments exhibiting motion behaviors dominated by the effects of complex wheel-ground interactions. Characterizing these interactions is crucial both from the immediate robot autonomy perspective (for motion prediction and control) as well as a long-term predictive maintenance and diagnostics perspective. An ideal solution entails capturing precise state measurements for decisions and controls, which is considerably difficult, especially in increasingly unstructured outdoor regimes of operations for these robots. In this milieu, a framework to identify pre-determined discrete modes of operation can considerably simplify the motion model identification process. To this end, we propose an interactive multiple model (IMM) based filtering framework to probabilistically identify predefined robot operation modes that could arise due to traversal in different terrains or loss of wheel traction.

Figures (6)

• Figure 1: An interactive multiple model estimation framework implemented by employing a bank of parallel extended Kalman filters (EKFs) for identifying motion modes that can be utilized for strategic control policy selection. The highlighted EKF signify the probabilistically identified nearest motion mode that leads to subsequent model selection.
• Figure 2: Numerous deterministic motion modes realized due to changes in available traction. (Top) Modes for large traction changes (LTC) for operation in asphalt, grass and crushed concrete. (Bottom) Modes for small traction changes (STC) arising due to loss of traction at one or several wheels. Traction losses are experimentally induced by wrapping wheels with polyethylene tarps resulting in the following combinations : 0-1-2-3-4 wheels skidding (total 16 combinations) with red wheels indicating covering with tarp.
• Figure 3: Singular mode identification on three sets of terrain where the algorithm is expected to stay stable in one of the operation modes. (a) Expected mode: Crushed concrete, (b) Expected mode: Grass, (c) Expected mode: Asphalt.
• Figure 4: Varying mode identification on three sets of terrain where algorithm is expected to switch real-time within the operation modes. Vertical dashed lines at $t = 10s$ and $t = 20s$ signify physical change in the operation mode (Crushed concrete - to Grass - to Asphalt).
• Figure 5: Singular mode identification on four sets of wheel slip modes where algorithm is expected to stay stable in one of the operation modes. (a) Expected mode : Four wheel skid, (b) Expected mode :Two Front wheel skid, (c) Expected mode : Two right wheel skid.