Dear fellow ROS developers,
We are glad to present OpEn (github): a code generation framework that allows the developer to formulate a parametric nonconvex optimization problem in Python (e.g., a nonlinear model predictive control or moving horizon estimation problem) and build a ROS package.
OpEn generates Rust code: Rust is a modern, fast, memory-safe systems programming language. It is ideal for embedded applications.
This is how it works:
Find more information here
Easy Installation
Installation is very easy. You first need to install Rust and clang. You can install OpEn using pip:
pip install opengen
It’s fast
It can be 1-2 orders of magnitude faster compared to sequential quadratic programming and interior point methods (reference 1 and 2)
Code example in Python
This is a short example to demonstrate that we can generate a
import opengen as og # <---- OpEn
import casadi.casadi as cs # symbolic math & differentiation
u = cs.SX.sym("u", 5) # decision variable
p = cs.SX.sym("p", 2) # parameter
phi = og.functions.rosenbrock(u, p) # Rosenbrock function
c = cs.vertcat(1.5 * u[0] - u[1],
cs.fmax(0.0, u[2] - u[3] + 0.1))
bounds = og.constraints.Ball2(None, 1.5)
problem = og.builder.Problem(u, p, phi) \ # minimize phi(u; p) wrt u
.with_penalty_constraints(c) \ # subject to: c(u; p) = 0
.with_constraints(bounds) # u in bounds
meta = og.config.OptimizerMeta() \
.with_optimizer_name("rosenbrock")
ros_config = og.config.RosConfiguration() \ # ROS configuration
.with_package_name("parametric_optimizer") \ # package name
.with_node_name("open_node") \ # node name
.with_rate(35) # rate (in Hz)
build_config = og.config.BuildConfiguration() \
.with_build_directory("my_optimizers") \
.with_ros(ros_config)
solver_config = og.config.SolverConfiguration() \ # solver config
.with_tolerance(1e-5) \
.with_delta_tolerance(1e-4) \
.with_penalty_weight_update_factor(5)
builder = og.builder.OpEnOptimizerBuilder(problem, meta,
build_config, solver_config)
builder.build() # Build optimizer & ROS package
Note the line .with_ros(ros_config) - this will generate a ROS package (it contains a README file with instructions on how to use it and a launch file).
Applications
OpEn has been used in a number of real-life applications of robotics such as:
- embedded nonlinear model predictive control for obstacle avoidance on a ground vehicle: that was the first application of the algorithm (there have been lots of developments sice)
-
fast nonlinear model predictive control on a micro aerial vehicle: runs NMPC at 20Hz, but can go even faster if necessary
- tutorial: an auto-generated ROS package for Husky: this is a step-by-step guide to building your own model predictive controller to navigate a Husky unmanned ground vehicle
We hope you find this useful!
If so, here’s a few things you can do for OpEn:
- Give us a
on github
-
with us on Discord! - Submit a feature request or
report (
read first) - Submit a pull request on github ( read first)