JMOAB_Autopilot_ROS

This is an autonomous drive package for rover with jmoab-ros, please check jmoab-ros repo here.

This package is designed to work side by side with jmoab-ros package to provide a feature of autonomous drive in many applications.

It could run with the real actual robot with JMOAB and also in simulation world in Gazebo as Software In The Loop (SITL).

Currently, it supports an autonomous drive as following.

• GPS Waypoints Navigation

GPS Waypoints Navigation

Hardware

To run autonomous drive outdoor with GPS envioronment, you will need

• Ublox F9P GPS with a setup here

• BNO055 9-axes orieintation sensor with a setup here

and you need to GPS as close to the wheel base axis, or pivot point when turning, and Compass/IMU on the middle of the cart.

Algorithms

You can check on nodes/gps_waypoints_navigation.py for the code, I will explain roughly how does it work..

So first you will need a waypoints file as shown in waypoints/mission.txt for example. The format is same as Ardupilot as QGC WPL 110. So you can make the mission file from APM Planner and click on Save WP. I found that MAVLink has some issue of lat/lon decimal points, it's not accurate when transfer using MAVLink command, so I don't recommend to use "Write" or "Read" button on GCS.

Once the bot has waypoints file it will store on as array of lat_target_list[], lon_targe_list[], then we gonna iterate with target_wp index counter to go to next watpoint. There are 5 step to make the bot go from current point to target waypoint.

1. Get the smallest angle to turn, and direction to turn.

2. Turn with constant sbus_steering speed, until it reahced the angke within threshold.

3. Get the distance how far it should go

4. Keep going to that point by using PID cross-track and heading controllers, if the bot is inside threshold of x_track_error_start then it will use PID heading, but if bot goes out of that threshold it will use PID cross-track. To calculate a related angle and distance variable, please check on two following images.

5. Once the bot reached the target waypoint within goal_dist_thresh threshold, it will iterate to next point and start with the step1 again.

Finally, if the robot can achieve all the point successfully, then it will just finish the mission and wait for user to switch back to auto mode again, as step 6.

Run

• On the bot

  • You will need to run gps, atcart, compass, RTK nodes as explained from jmoab-ros doc.

  • On this package you will need to run rosrun jmoab_autopilot_ros gps_waypoints_server.py to start the parameters server. This is need when you firstly want to tune up the bot.

  • And on jmoab_autopilot/nodes you need to run a navigation script as python gps_waypoints_navigation.py, you can copy this script to your project directory and make some modification according to what you need.

• On the PC

  • Then you can run rosrun rqt_reconfigure rqt_reconfigure on you PC for GUI adjuster.

Parameters

We are using ros_dynamic_reconfigure, so we can adjust the paramters lively from rqt_reconfigure GUI. Once you open it up it's like this,

The paramters in here are same as cfg/GpsWaypoints.yaml, so once you satisfied with tuning, you will need to update it manually on the bot on this file too (I will find the way for automate this later).

Steering

• max_start_str: 1054, this is a maximum sbus steering to make the cart start to turn right, each cart has different value depends on load

• min_start_str: 986, this is a minimum sbus steering to make the cart start to turn left, each cart has different value depends on load

• str_adj: 100, this is an adjuster sbus value to use with output PID mapping, so max output pid will use this full range +/- max/min_start_str

• skid_adj: 80, this is a pivot turn adjuster sbus value, so it will be +/- on max/min_start_str for right/left pivot-turning

• str_mid: 969

Throttle

• max_start_thr: 1080, this is a maximum sbus throttle to make the cart start to move forward.

• thr_adj: 100, this is an adjuster sbus value to + on max_start_thr, to use as constant speed during navigation.

• thr_slowest: 1140, this is the slowest sbus throttle in case of short distance navigation, we don't want the bot to move too fast when short range.

• thr_mid: 1024, a middle sbus throttle, no need to change this.

General Judgement Threshold

• goal_dist_thresh: 0.2, this is a radius of waypoints, it means if the cart is closed to target waypoint within this radius, it's considered as reached the point.

• goal_ang_thresh: 1.0, this is a threshold angle when cart is turning as pivot, so if difference between goal target angle and current angle is under this, it's considered as arrival at goal target angle.

• x_track_error_start: 0.12, this is a cross-track distance from the route to make the PID cross-track controller starts, when x_track_error less than this, the cart will use PID heading controller instead.

• x_track_repose_dist: 0.40, this is the maximum cross-track distance that tell the cart is too far from the route and it will start re-target heading again.

• pid_hdg_out_thresh: 0.7, if diff_hdg is less than this, it won't use PID heading and just keep driving straight.

• pid_x_out_thresh: 0.1, if x_track_error is less than this, it won't use PID cross-track, it's better to make it same as x_track_error_start or not higher.

PID gain

• hdg_p: 40.0, P gain of heading control

• hdg_i: 0.0, I gain of heading control

• hdg_d: 0.0, D gain of heading control

• cross_p: 150.0, P gain of cross-track control

• cross_i: 10.0, I gain of cross-track control

• cross_d: 0.0, D gain of cross-track control

Tips

• If the heading is not good, the bot will move like snake and has a hard time to go back to the route, so please calibrate the heading offset again see here.

Greenhouse Navigation

The greenhouse navigation is using only lidar and IMU to move around in the greenhouse.

The navigation pattern is using the a step sequence below

1. Wall-following: where the bot is using either left or right wall distance to measure how far from itself and try to keep the distance

2. Lane-changing: where the bot is outside the lane and try to change to another lane

3. Re-heading: when the bot arrived on the new lane, it will adjust the heading again

4. Wall-following(2): this is similar as first step

5. U-turning: where the bot front_stop_dist found some wall all of obstacle, it will turn around and count as one round completed.

The bot will follow this pattern forever, so for higher application you will need to check on shelf_counter inside the code, and stop whenever the last shelf of greenhouse got completed.

The parameters are showing,