Self-Assessment of Evidential Grid Map Fusion for Robust Motion Planning

TL;DR

This work tackles the challenge of conflicting sensor measurements in autonomous perception by leveraging Subjective Logic to classify grid-cell conflicts within an evidential grid map. It introduces a self-assessment framework that computes a degradation score $\alpha$ to quantify the impact of conflicts and guide decision-making, including when sensor calibration issues arise. The authors then propagate conflict information to a conflict-aware path planning module (based on Hybrid A* with dilation-driven costs), enabling robust planning that can either avoid or cautiously navigate through conflicting regions. The approach demonstrates improved resilience to degraded environment representations and reduces unnecessary replanning while maintaining safety in autonomous motion planning.

Abstract

Conflicting sensor measurements pose a huge problem for the environment representation of an autonomous robot. Therefore, in this paper, we address the self-assessment of an evidential grid map in which data from conflicting LiDAR sensor measurements are fused, followed by methods for robust motion planning under these circumstances. First, conflicting measurements aggregated in Subjective-Logic-based evidential grid maps are classified. Then, a self-assessment framework evaluates these conflicts and estimates their severity for the overall system by calculating a degradation score. This enables the detection of calibration errors and insufficient sensor setups. In contrast to other motion planning approaches, the information gained from the evidential grid maps is further used inside our proposed path-planning algorithm. Here, the impact of conflicting measurements on the current motion plan is evaluated, and a robust and curious path-planning strategy is derived to plan paths under the influence of conflicting data. This ensures that the system integrity is maintained in severely degraded environment representations which can prevent the unnecessary abortion of planning tasks.

Figures (13)

• Figure 1: Sensor measurements are fused in the evidential grid map whose cells are then categorized in the conflict-aware classification module. They are evaluated in the self-assessment module and assigned movement costs or set to non-drivable depending on their distance to the ego vehicle. This leads to conflict-aware path planning on the categorized grid map, which retains information about conflicting poses in the path for underlying stages of motion planning. Last, the categorized grid map is used to calculate a degradation score, which evaluates the overall integrity of the sensor data.
• Figure 2: (a) Categorization into occupied, unknown and free cells. (b) Categorization into drivable and non-drivable cells.
• Figure 3: A binomial opinion $\omega_X$ is illustrated in a barycentric triangle. The three axes of belief, disbelief, and uncertainty are represented by $b_X$, $d_X$, and $u_X$, respectively, and $a_X$ is the prior projecting $\omega_X$ to $P_X(x)$.
• Figure 4: The cumulative fusion of different opinions (dots) with their result (crosses) is illustrated in a barycentric triangle. The red dots are conflicting opinions; the blue dots are supporting ones. The fusion of an opinion with a neutral opinion with $u_X=1.0$ is shown in orange.
• Figure 5: Point clouds of two LiDAR sensors mounted with a distance of 1 on top of a vehicle in the Carla simulator at a parking lot in Town04. The colors encode different object categories. One of the sensors has a rotational calibration error of 5° around its z-axis. The effect of the calibration error can, in particular, be observed on the parked cars in green and the street light poles in yellow (marked by the red rectangle).

Theorems & Definitions (2)

• Definition 1: Probabilistic Multi-Sensor Fusion thrun2005probabilistic
• Definition 2: Multinomial Opinion josang2016