Proactive Anomaly Detection for Robot Navigation With Multi-Sensor Fusion

Abstract

Despite the rapid advancement of navigation algorithms, mobile robots often produce anomalous behaviors that can lead to navigation failures. The ability to detect such anomalous behaviors is a key component in modern robots to achieve high-levels of autonomy. Reactive anomaly detection methods identify anomalous task executions based on the current robot state and thus lack the ability to alert the robot before an actual failure occurs. Such an alert delay is undesirable due to the potential damage to both the robot and the surrounding objects. We propose a proactive anomaly detection network (PAAD) for robot navigation in unstructured and uncertain environments. PAAD predicts the probability of future failure based on the planned motions from the predictive controller and the current observation from the perception module. Multi-sensor signals are fused effectively to provide robust anomaly detection in the presence of sensor occlusion as seen in field environments. Our experiments on field robot data demonstrates superior failure identification performance than previous methods, and that our model can capture anomalous behaviors in real-time while maintaining a low false detection rate in cluttered fields.Code is available at https://github.com/tianchenji/PAAD.

A Day/Night Leader-Following Method Based on Adaptive Federated Filter for Quadruped Robots

“…N. Dwek et al [ 26 ] designed sensor fusion and outlier detection to improve robust Ultra Wide Band positioning performance. T. Ji et al [ 27 ] used fault detection for navigation failure in agricultural environments to combat sensor occlusion in cluttered environments.…”

Autoencoding a Soft Touch to Learn Grasping from On‐Land to Underwater

“…The recent adoption of a SVAE model for identifying critical underlying factors for prediction demonstrates the promising potential for application in robotic grasping. [33] This study investigates the transferability of grasping knowledge from on-land to underwater via a vision-based soft robotic finger that learns 6D forces and torques (FT) using SVAE. Using real-time images collected from an in-finger camera that captures the soft finger's whole-body deformations while interacting with physical objects on-land and underwater, we established a learning-based approach to introduce tactile intelligence for soft, delicate, and reactive grasping underwater, making it a promising solution to support scientific discoveries in interdisciplinary research.…”

“…The recent adoption of a SVAE model for identifying critical underlying factors for prediction demonstrates the promising potential for application in robotic grasping. [ 33 ]…”

Traversing Supervisor Problem: An Approximately Optimal Approach to Multi-Robot Assistance

“…Although recent research efforts have made noteworthy progress on developing trustworthy robot autonomy [16,34,40], the deployment of low-cost robots in real-world environments has shown that they usually face difficulties in completing the tasks independently [32]. For example, due to the environmental complexity and terrain variability in agricultural environments, compact field robots (Figure 1a) deployed between rows of crops may fail the navigation task and get stuck in error states (Figure 1b), in which physical assistance from a human supervisor is required to continue the robot task [14,15]. To ensure the smooth operation of such multi-robot systems, the supervision and management of robot fleets are necessary in the presence of imperfect autonomy.…”