TL;DR
This paper presents a novel inflatable soft robotic arm designed for safe human-robot interaction and pick-and-place tasks. By leveraging advanced fabrication and a sophisticated control strategy that accounts for joint stiffness and payload, the arm achieves robust manipulation and safe collaboration, even with human contact.
What Problem Does This Paper Address?
The increasing desire for robots to work alongside humans, particularly in collaborative settings, necessitates robots that are inherently safe. Traditional rigid robots pose a risk of injury during unexpected collisions. Soft robots, with their compliant nature and lightweight construction, offer a promising solution. However, designing and controlling these systems presents unique challenges, including complex dynamics, difficulty in precise modeling, and ensuring reliable manipulation. This paper addresses these challenges by developing a robust, inflatable soft robotic arm that prioritizes safety while enabling sophisticated pick-and-place operations, including safe handover to humans.
Methods
The authors developed an inflatable robotic arm using fabric and three bellow-type pneumatic actuators. Key methodological contributions include:
- Safety-Driven Design: A safety analysis based on first principles guided the design, emphasizing a lightweight structure to minimize collision forces.
- Advanced Fabrication: State-of-the-art manufacturing techniques like high-frequency welding were employed to ensure robustness and reliability.
- Control Allocation Strategy: A novel control allocation strategy was developed to decouple rotational degrees of freedom from joint stiffness, allowing for independent control. This strategy also ensures balanced usage of actuators and valves.
- Linear Parameter-Varying (LPV) Modeling: A dynamic model was derived that accounts for the changing joint stiffness (controlled by lower actuator pressure) and the presence of a payload mass. This model is crucial for precise control.
- Cascaded Control Architecture: A two-level control system was implemented. An inner loop controls actuator pressures at 1000 Hz, while an outer loop controls the arm's angles at 50 Hz, exploiting a timescale separation between pressure and arm dynamics.
- Feed-Forward Compensation: Closed-form feed-forward strategies were developed to compensate for disturbances caused by gravity and longitudinal elongation during grasping.
- Valve Flow Capacity Analysis: The critical role of valve flow capacity in maintaining timescale separation and preventing oscillatory behavior was investigated.
Datasets
Not discussed in the paper. The authors performed system identification experiments to derive model parameters.
Key Insights and Ideas
- Lightweight Design is Paramount for Safety: The paper strongly emphasizes that reducing the robot arm's mass is the most effective way to minimize interaction forces during collisions, a key insight for safe human-robot collaboration.
- Joint Stiffness and Elongation are Coupled: A crucial finding is that increasing joint stiffness through simultaneous actuator inflation also leads to longitudinal elongation of the soft joint. This coupling is leveraged for grasping but cannot be independently controlled.
- Valve Flow Capacity is Critical for Control Performance: Insufficient valve flow capacity can lead to a breakdown of the assumed timescale separation in cascaded control, resulting in oscillatory behavior and degraded performance. This highlights the importance of selecting appropriate pneumatic components.
- Control Allocation Simplifies Modeling and Control: The proposed control allocation strategy effectively decouples rotational control from stiffness control, making the overall system more manageable and enabling independent manipulation of these aspects.
- Softness Enables Novel Interaction Detection: The intrinsic compliance of the soft robotic arm is leveraged to detect human interaction, allowing for a safe handover by sensing when the object is moved by the human.
Key Results
- The developed inflatable robotic arm demonstrated robust pick-and-place capabilities, including grasping objects up to 0.6 kg.
- Two distinct applications were successfully realized: a pick-and-place task to a known location and a collaborative pick-and-handover to a human.
- The feed-forward control strategy significantly reduced tracking errors during longitudinal elongation for grasping, improving performance by a factor of two.
- System identification experiments validated the LPV model's accuracy in predicting the arm's dynamics under varying joint stiffness and payload conditions.
- The analysis of valve flow capacity provided a tool for designers to select appropriate valves to ensure proper timescale separation for stable cascaded control.
Limitations and Future Work
- Viscoelastic Material Behavior: The paper acknowledges that the linear parameter-varying model does not fully capture the viscoelastic effects of the soft materials, which can lead to stress relaxation and slow dynamic shifts. Future work could incorporate disturbance estimation techniques to address these slower dynamics.
- Limited Degrees of Freedom: The current system's degrees of freedom might be limiting for more complex manipulation tasks. The authors suggest connecting the inflatable arm to a conventionally actuated base to increase maneuverability while maintaining soft components.
- Interaction Force Estimation: The current interaction detection relies on positional deviation. Future work could focus on estimating interaction forces directly from the model for enhanced safety.
- More Complex Applications: The paper suggests exploring applications beyond pick-and-place, such as joint object manipulation with humans, leveraging the adjustable joint stiffness.
Summary updated