Impact-Aware Robotic Manipulation: Quantifying the Sim-To-Real Gap for Velocity Jumps

Jari van Steen,Daan Stokbroekx,Nathan van de Wouw,Alessandro Saccon

2024-11-10

Abstract:Impact-aware robotic manipulation benefits from an accurate map from ante-impact to post-impact velocity signals to support, e.g., motion planning and control. This work proposes an approach to generate and experimentally validate such impact maps from simulations with a physics engine, allowing to model impact scenarios of arbitrarily large complexity. This impact map captures the velocity jump assuming an instantaneous contact transition between rigid objects, neglecting the nearly instantaneous contact transition and impact-induced vibrations. Feedback control, which is required for complex impact scenarios, will affect velocity signals when these vibrations are still active, making an evaluation solely based on velocity signals as in previous works unreliable. Instead, the proposed validation approach uses the reference spreading control framework, which aims to reduce peaks and jumps in the control feedback signals by using a reference consistent with the rigid impact map together with a suitable control scheme. Based on the key idea that selecting the correct rigid impact map in this reference spreading framework will minimize the net feedback signal, the rigid impact map is experimentally determined and compared with the impact map obtained from simulation, resulting in a 3.1% average error between the post-impact velocity identified from simulations and from experiments.

Robotics

What problem does this paper attempt to address?

The core problem that this paper attempts to solve is: **How to accurately generate and verify the impact map between the robot and the environment, especially in complex multi - body collision scenarios**. Specifically, the author focuses on the speed jump problem from simulation to practical application (Sim - to - Real Gap) and proposes an experimental verification method based on the reference - spreading control framework.

### Main Research Questions

1. **Generation of Impact Map**:
- The author proposes a method of using a physics engine to generate an impact map, which can describe the speed changes before and after the collision. This map is very important for supporting the motion planning and control of robots.

2. **Quantification of Sim - to - Real Gap**:
- Due to vibrations and other unmodeled factors in the real world, directly applying simulation results to actual systems may lead to errors. Therefore, the author attempts to quantify and reduce this gap to ensure the reliability of simulation results in practical applications.

3. **Handling of Complex Impact Scenarios**:
- For complex collisions between multiple objects occurring simultaneously (such as two - arm grasping), traditional analysis models are difficult to apply. The method proposed by the author can handle these complex scenarios and verify its effectiveness through experiments.

4. **Influence of Feedback Control**:
- In complex impact scenarios, the feedback control will affect the speed signal when the vibration caused by the impact still exists, making the evaluation based only on the speed signal unreliable. For this reason, the author introduces the reference - spreading control framework to minimize the control feedback signal, thereby more accurately evaluating the effectiveness of the impact map.

### Overview of Solutions

1. **Impact Map Generation**:
- Use a physics engine (such as AGX Dynamics) to simulate collision events and generate a rigid impact map assuming instantaneous contact transition. This map does not consider approximate instantaneous contact transition and vibrations caused by impacts.

2. **Experimental Verification**:
- Conduct experimental verification through the reference - spreading control framework, which aims to minimize the net feedback signal by selecting the correct rigid impact map. The experimental results show that the average error between the post - impact speed predicted by the simulation and the experimental results is 3.1%.

3. **Control Strategy**:
- Introduce reference - spreading control (Reference Spreading Control, RS) to reduce peaks and jumps in the control input, especially during and after impacts. The RS control framework includes different control modes before the impact, during the transition period, and after the impact to ensure a smooth transition throughout the process.

Through these methods, the author has successfully solved the problems of modeling and verification in complex impact scenarios between robots and the environment, providing strong support for the development of impact - aware robotics technology.