Register now After registration you will be able to apply for this opportunity online.
Advanced Co-Design Framework for Legged Robots
This project seeks to advance the field of legged robotics by creating a versatile and accessible co-design framework that integrates mechanical design and control optimization.
Legged robots present unique challenges in design and control due to their complex dynamics and the need for energy-efficient, agile locomotion. Some recent work proved this approach viable for these systems [1-4]. This project aims to develop an advanced co-design framework that simultaneously optimizes the mechanical structure and control strategies of legged robots, addressing key challenges in the field.
**Objectives**
Develop a versatile co-design framework that integrates mechanical design and control optimization for legged robots.
Enhance the dynamic performance of legged robots through simultaneous hardware and software optimization.
Improve the accessibility of legged robot design for users with varying levels of expertise.
Bridge the gap between simulation and real-world performance in legged robotics.
**Key components of this project**
_Optimization/Learning-based Approach_:
The developed framework should achieve the selection of efficient and effective designs for several robot motion tasks. The approach will:
- Optimize both the robot's mechanical structure and its dynamic maneuvers concurrently.
- Possibly utilize gradient information and differentiable simulation [5] to improve the convergence capability. Or alternative explore the power of reinforcement learning for the generation of policies.
- Develop efficient computation methods for necessary design changes to maintain an interactive design flow.
- Use adaptive control algorithms: this allows the design of control strategies that can adapt to varying robot morphologies and environmental conditions.
_Interactive Design Interface_:
To make the co-design process more accessible, an interactive computational design system could be employed:
- Enabling users to specify high-level descriptions of desired robot morphologies and behaviors, such as in [7, 8].
- Allowing the user to automatically generate motion plans, similarly to [6].
- Optimizing robot size, motion duration, and actuator selection.
- Providing feedback on design feasibility and performance.
_Validation and Testing_: In the final stage of the work, the project could lead to experiments:
- Simulation Studies: Conduct extensive simulations to validate the co-design framework across various legged robot configurations and tasks.
- Hardware Prototyping: Develop physical prototypes of optimized designs to verify real-world performance.
- Sim-to-Real Transfer: Evaluate the transferability of optimized designs and control policies from simulation to physical robots.
**References:**
[1] “Co-designing versatile quadruped robots for dynamic and energy-efficient motions”, G. Fadini et al., Robotica, 2024.
[2] “Skaterbots: optimization-based design and motion synthesis for robotic creatures with legs and wheels”. Moritz Geilinger et al., ACM Trans. Graph., 2018.
[3] “Computational design of robotic devices from high-level motion specifications”. Sehoon Ha et al. IEEE Trans. Robotics 34.5, 2018.
[4] “Control-aware design optimization for bioinspired quadruped robots”. De Vincenti Flavio, Kang Dongho, and Coros Stelian. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2021.
[5] “ADD: Analytically differentiable dynamics for multi-body systems with frictional contact”. Moritz Geilinger et al. ACM Trans. Graph. 39.6, 2020.
[6] “Predictive Sampling: Real-time Behaviour Synthesis with MuJoCo.” Taylor et al. Howell. https://arxiv.org/abs/2212.00541, Dec 2022.
[7] “Interactive design of animated plushies”. M. Bern James, Chang Kai-Hung, and Stelian Coros. ACM Trans. Graph. 36.4, 2017.
[8] “Joint optimization of robot design and motion parameters using the implicit function theorem”. Sehoon Ha et al. Robotics: Science and Systems XIII, 2017.
[9] “Chacra: An interactive design system for rapid character crafting”. Vittorio Megaro et al., The Eurographics / ACM SIGGRAPH Symposium on Computer Animation, SCA 2014, Copenhagen, Denmark, 2014.
Legged robots present unique challenges in design and control due to their complex dynamics and the need for energy-efficient, agile locomotion. Some recent work proved this approach viable for these systems [1-4]. This project aims to develop an advanced co-design framework that simultaneously optimizes the mechanical structure and control strategies of legged robots, addressing key challenges in the field.
**Objectives**
Develop a versatile co-design framework that integrates mechanical design and control optimization for legged robots. Enhance the dynamic performance of legged robots through simultaneous hardware and software optimization. Improve the accessibility of legged robot design for users with varying levels of expertise. Bridge the gap between simulation and real-world performance in legged robotics.
**Key components of this project**
_Optimization/Learning-based Approach_: The developed framework should achieve the selection of efficient and effective designs for several robot motion tasks. The approach will:
- Optimize both the robot's mechanical structure and its dynamic maneuvers concurrently. - Possibly utilize gradient information and differentiable simulation [5] to improve the convergence capability. Or alternative explore the power of reinforcement learning for the generation of policies. - Develop efficient computation methods for necessary design changes to maintain an interactive design flow. - Use adaptive control algorithms: this allows the design of control strategies that can adapt to varying robot morphologies and environmental conditions.
_Interactive Design Interface_: To make the co-design process more accessible, an interactive computational design system could be employed:
- Enabling users to specify high-level descriptions of desired robot morphologies and behaviors, such as in [7, 8]. - Allowing the user to automatically generate motion plans, similarly to [6]. - Optimizing robot size, motion duration, and actuator selection. - Providing feedback on design feasibility and performance.
_Validation and Testing_: In the final stage of the work, the project could lead to experiments:
- Simulation Studies: Conduct extensive simulations to validate the co-design framework across various legged robot configurations and tasks.
- Hardware Prototyping: Develop physical prototypes of optimized designs to verify real-world performance.
- Sim-to-Real Transfer: Evaluate the transferability of optimized designs and control policies from simulation to physical robots.
**References:** [1] “Co-designing versatile quadruped robots for dynamic and energy-efficient motions”, G. Fadini et al., Robotica, 2024.
[2] “Skaterbots: optimization-based design and motion synthesis for robotic creatures with legs and wheels”. Moritz Geilinger et al., ACM Trans. Graph., 2018.
[3] “Computational design of robotic devices from high-level motion specifications”. Sehoon Ha et al. IEEE Trans. Robotics 34.5, 2018.
[4] “Control-aware design optimization for bioinspired quadruped robots”. De Vincenti Flavio, Kang Dongho, and Coros Stelian. IEEE/RSJ International Conference on Intelligent Robots and Systems, 2021.
[5] “ADD: Analytically differentiable dynamics for multi-body systems with frictional contact”. Moritz Geilinger et al. ACM Trans. Graph. 39.6, 2020.
[6] “Predictive Sampling: Real-time Behaviour Synthesis with MuJoCo.” Taylor et al. Howell. https://arxiv.org/abs/2212.00541, Dec 2022.
[7] “Interactive design of animated plushies”. M. Bern James, Chang Kai-Hung, and Stelian Coros. ACM Trans. Graph. 36.4, 2017.
[8] “Joint optimization of robot design and motion parameters using the implicit function theorem”. Sehoon Ha et al. Robotics: Science and Systems XIII, 2017.
[9] “Chacra: An interactive design system for rapid character crafting”. Vittorio Megaro et al., The Eurographics / ACM SIGGRAPH Symposium on Computer Animation, SCA 2014, Copenhagen, Denmark, 2014.
**Expected Outcomes**
1. Development of a versatile co-design framework applicable to various legged robot types and tasks.
2. Significant improvements in energy efficiency and dynamic performance of legged robots.
3. Extending the accessibility of legged robot design to non-expert users.
4. Better understanding of the relationship between hardware design and robot behavior.
**Expected Outcomes**
1. Development of a versatile co-design framework applicable to various legged robot types and tasks.
2. Significant improvements in energy efficiency and dynamic performance of legged robots.
3. Extending the accessibility of legged robot design to non-expert users.
4. Better understanding of the relationship between hardware design and robot behavior.
Master projects - CoDesign.pdf
