Standing Tall: Sim to Real Fall Classification and Lead Time Prediction for Bipedal Robots
Gokul Prabhakaran, J.W Grizzle, M. Eva Mungai
AI summary
Problem
Bipedal robots struggle with stability in unstructured environments, yet existing fall prediction methods lack real-time capability, adequate lead time, or robustness to diverse physical faults on full-sized humanoids.
Approach
The authors adapt a 1D CNN-based prediction framework to run online in real-time on the Digit robot, validating it across simulation and hardware while analyzing and improving its robustness to omnidirectional disturbances.
Key results
- Real-time implementation matches offline accuracy with zero false positives
- Achieves 1.1s average lead time and 0.03s max lead time error
- Demonstrates 0.97 recovery rate in simulation when triggering recovery controllers
- Fine-tuning strategy reduces average false positive rate by 0.05 and max lead time error by 1.19s
Why it matters
Provides a validated, real-time safety framework that enables full-sized humanoid robots to reliably detect falls and execute recovery strategies in dynamic, real-world settings.
Abstract
This paper extends a previously proposed fall prediction algorithm to a real-time (online) setting, with im- plementations in both hardware and simulation. The system is validated on the full-sized bipedal robot Digit, where the real- time version achieves performance comparable to the offline implementation while maintaining a zero false positive rate, an average lead time (defined as the difference between the true and predicted fall time) of 1.1s (well above the required minimum of 0.2s), and a maximum lead time error of just 0.03s. It also achieves a high recovery rate of 0.97, demonstrating its effectiveness in real-world deployment. In addition to the real- time implementation, this work identifies key limitations of the original algorithm, particularly under omnidirectional faults, and introduces a fine-tuned strategy to improve robustness. The enhanced algorithm shows measurable improvements across all evaluated metrics, including a 0.05 reduction in average false positive rate and a 1.19s decrease in the maximum error of the average predicted lead time.