Control for the inverted pendulum with Reinforcement Learning -- 2
Budget: €1,500 – €3,000 EUR
The company Lucas-Nülle develops training systems for technical training and studies. We recently added machine learning to our program: https://youtu.be/Uq2GNNDWRIo. So far we have limited ourselves to computer vision, i.e. classification and object detection. We use the Jetson Xavier NX with Tensorflow 1.15 (Python 3.6.9) for this.
In the next step, we want to connect the Jetson Xavier NX to our digital controller via the CAN bus and in this way control our inverted pendulum. This already works with a cascade controller and a state controller: https://youtu.be/qhdXtC-J6ks. Now the system is to be regulated with an AI controller and we are currently looking for an expert in the field to support us with this.
The tasks are:
• Connection of the Jetson Xavier NX with our digital controller using the PCAN-USB adapter (https://www.peak-system.com/PCAN-USB.199.0.html), for the exchange of control and manipulated variables
• Creation of a real-time position and angle control of the pendulum based on reinforcement learning
• The quality of the control result should roughly correspond to the result from the video linked above, i.e. the control should be "calm" and disturbances should be corrected
• Complete documentation of the designing process
In the next step, we want to connect the Jetson Xavier NX to our digital controller via the CAN bus and in this way control our inverted pendulum. This already works with a cascade controller and a state controller: https://youtu.be/qhdXtC-J6ks. Now the system is to be regulated with an AI controller and we are currently looking for an expert in the field to support us with this.
The tasks are:
• Connection of the Jetson Xavier NX with our digital controller using the PCAN-USB adapter (https://www.peak-system.com/PCAN-USB.199.0.html), for the exchange of control and manipulated variables
• Creation of a real-time position and angle control of the pendulum based on reinforcement learning
• The quality of the control result should roughly correspond to the result from the video linked above, i.e. the control should be "calm" and disturbances should be corrected
• Complete documentation of the designing process