Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/cryscan/to-ihc-2
Trajectory optimization with implicit hard constraints.
https://github.com/cryscan/to-ihc-2
Last synced: about 1 month ago
JSON representation
Trajectory optimization with implicit hard constraints.
- Host: GitHub
- URL: https://github.com/cryscan/to-ihc-2
- Owner: cryscan
- Created: 2021-08-18T03:52:22.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2022-01-22T08:17:52.000Z (almost 3 years ago)
- Last Synced: 2024-10-20T11:43:21.687Z (2 months ago)
- Language: C++
- Homepage:
- Size: 411 KB
- Stars: 1
- Watchers: 2
- Forks: 0
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Trajectory Optimization with Hard Contacts
This is an unofficial implementation of [1]. The algorithm makes use of concepts from bi-level optimization to find
gradients of the system dynamics including the constraint forces and subsequently solve the optimal control problem with
the unconstrained iLQR algorithm.
The moving picture shows a trajectory generated.
## Implementation
The robot used here has nearly the same configurations as in [1], which is a planer hopper with 4 degree of freedoms.
The robot's kinetic and dynamics models are generated using [`RobCoGen`](https://robcogenteam.bitbucket.io/).
The time stepping scheme in [1] has some problem where the end effector is easy to penetrate the ground, which
significantly hinders the optimization. Some fixes are taken place according to [2].
To improve the stability of iLQR iterations, regularization and line-search schemes in [3] are used.
## Running
This project is written in `c++` and is only tested under Linux.
### Requirements
The code depends on the linear algebra library [`Eigen3`](https://eigen.tuxfamily.org).
[`RobCoGen`](https://robcogenteam.bitbucket.io/) headers are required to be installed, even if not switching into other
robot models.
The code also depends on [`CppAD`](https://github.com/coin-or/CppAD.git)
and [`CppADCodeGen`](https://github.com/joaoleal/CppADCodeGen.git). These are both `c++` header-only libraries that
support fast auto-differentiation.
### Clone and Compile
#### Clone the repository
```shell
$ git clone https://github.com/cryscan/to-ihc-2.git
$ cd to-ihc-2
```
#### Build the project
```shell
$ mkdir build
$ cd build
$ cmake ..
$ make
```
#### Run the code
To run the code, create a `in.txt` under the `build` folder with the following contents:
```
# Iterations and minimum feedforward gain
10 1e-4
# Initial configuration
0.5 0.0 0.785 -1.57 0.0 0.0 0.0 0.0
# Target configuration
-0.175 1.0 0.785 -1.57 0.0 0.0 0.0 0.0
# Running cost
0.01 0.0 0.02 0.02 0.0 0.0 0.04 0.04
0.001 0.001
# Final cost
800.0 2000.0 400.0 400.0 200.0 200.0 200.0 200.0
0.0 0.0
# Horizon
200
# initial PD gains
0 0 40.0 40.0 0 0 1.0 1.0
```
Then run
```shell
$ ./control
```
In the first run it generates codes for dynamics, cost, etc. and compiles them, which may take some time. It won't
compile again in future runs if the compiled dynamic libraries exist.
Very quickly the result trajectory will be stored in `out.txt`. It can be visualized by running
```shell
$ python3 ../visual.py
```
Alternatively, running
```shell
$ ./simulation
```
generates a rollout without any control signals.
# Reference
1. J. Carius, R. Ranftl, V. Koltun and M. Hutter, "Trajectory Optimization With Implicit Hard Contacts," in IEEE
Robotics and Automation Letters, vol. 3, no. 4, pp. 3316-3323, Oct. 2018, doi: 10.1109/LRA.2018.2852785.
2. J. Carius, R. Ranftl, V. Koltun and M. Hutter, "Trajectory Optimization for Legged Robots With Slipping Motions," in
IEEE Robotics and Automation Letters, vol. 4, no. 3, pp. 3013-3020, July 2019, doi: 10.1109/LRA.2019.2923967.
3. Y. Tassa, T. Erez and E. Todorov, "Synthesis and stabilization of complex behaviors through online trajectory
optimization," 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2012, pp. 4906-4913, doi:
10.1109/IROS.2012.6386025.