Trajectory Stabilization with AcroMonk

The following instructions assume that the scripts will be executed on the Raspberry Pi. The required packages that need to be installed on the Raspberry Pi are provided in the requirement.txt.

A Makefile is provided in the following folder for the automatic execution of the robust forward brachiation using TVLQR and PD controller:

software/python/real_system/Makefile

The following parameters can be used for arbitrary behavior:

TEST          : stabilizing controller (pd, tvlqr), default=pd
FILE          : swing trajectory (BF trajectory in data/trajectories/closed_loop folder), default=BF_tvlqr.csv
MOTOR_ID      : mjbots servo id, default=7
BUS_NUMBER    : bus number on pi3hat, default=4
FILE_RECOVERY : recovery trajectory (Recovery trajectory in data/trajectories/closed_loop folder), default=ZF_tvlqr.csv
INIT_CONF     : initial configuration of the AcroMonk (1, 2), default=1: description provided in initial configuration section of this README
NBRACH        : number of brachiation maneuvers, default=1

for example:

make NBRACH=3  TEST=tvlqr

executes 3 brachiation motions with TVLQR stabilization. After typing the execution command on the terminal, the initialization of the parameters and IMU will be started. If it returns no error in the terminal, you need to type in y if ready for the execution on the real system. The plots and the csv file are saved in the results/real_system folder.

Initial configuration

The zero offset coordinate of the motor will be set when the motor is energized. It is of great importance that the energization takes place when the robot has a hanging arm and $\theta_2=0$. The following table provides an overview of the possible initialization of the robot for the execution of the script. The arm with the computing compartment, i.e. Raspberry Pi and IMU, is shown in green.

Here is the main difference between configurations:

• INIT_CONF=1: IMU swings with the arm and the motor axis is outward, i.e. positive rotation is counter-clockwise.
• INIT_CONF=2: IMU arm is attached to the bar and the motor axis is pointing towards the page, i.e. positive rotation is clockwise.

Robust Forward Brachiation with AcroMonk

For the robust forward brachiation, if you start the script's execution with zero configuration and then press y, the state machine detects the hanging position of the arm and starts with the recovery mode ZF and further continues with the rest of the brachiation maneuvers. However, if you place the robot in both arms attached configuration after the initialization phase from zero configuration, and then press y, the execution will be started from the BF maneuver and continues with the brachiation.

Atomic Behavior with AcroMonk

The following instructions allow the execution of the atomic behaviors ZF, BF, ZB, FB on the real system:

BF

Place the robot in the initial position where both arms are attached to the bars and type one of the following commands depending on the INIT_CONF:

make 

or

make NBRACH=2 INIT_CONF=2

ZF

Place the robot in the hanging configuration and type one of the following commands depending on the INIT_CONF:

make

or

make NBRACH=2 INIT_CONF=2

ZB

Place the robot in the hanging configuration and type one of the following commands depending on the INIT_CONF:

make FILE_RECOVERY=ZB_pd.csv

or

make NBRACH=2 INIT_CONF=2 FILE_RECOVERY=ZB_pd.csv

FB

To execute the FB maneuver, you need to execute the script in the following folder software/python/real_system/FB_PD/tests/FB_brachiation.py.

Notes and cautions

• This script works only with the INIT_CONF=1
• The following steps need to be taken for the real system test:
  • Place the robot in the hanging zero configuration
  • Type in the terminal: sudo -E python3 FB_brachiation.py and if it returns no error, type in y and press Enter
  • The script waits for 10 seconds to let the operator attach the IMU arm to the front bar

Forward Brachiation using Reinforcement Learning

A guide is provided for the execution of the Reinforcement learning algorithm on the AcroMonk.