mit-racecar / racecar Goto Github PK

Is there any documentation for the ackermann_cmd_mux package? When I fire up the Gazebo simulation I can see a complex graph (rqt_graph). I understand that this package reads inputs from different sources (joystick, safety controller and autonomous algorithms like the wall follower) and decide which one to use based on the joystick inputs and the priorities. The priorities for the topics are in the following order:

1. /vesc/ackermann_cmd_mux/input/teleop (highest priority)
2. /vesc/ackermann_cmd_mux/input/safety
3. /vesc/ackermann_cmd_mux/input/navigation (lowest priority)

My understanding is based on this presentation - Racecar Basics

My question

If I have to use an autonomous algorithm along with using a joystick then to which topic should the autonomous algorithm publish control commands to? Currently the joystick is publishing to /vesc/low_level/ackermann_cmd_mux/input/teleop. I have tried /vesc/ackermann_cmd_mux/input/navigation and /vesc/high_level/ackermann_cmd_mux/input/nav_0, none worked.

Background

I am right now testing a "wall following algorithm" in the F1/10 simulator. My node reads the laser scans, computes the PID control values to make the car go parallel to the wall and publishes a AckermannDriveStamped message on the topic /vesc/ackermann_cmd_mux/input/teleop. It works fine.

Now I want to be able to use the joystick, just in case if I have to course correct, while my wall following algorithm is running.

So I understand that my wall follower node cannot use the /vesc/ackermann_cmd_mux/input/teleop topic anymore since that would mean that the algorithm commands would have the highest priority. So I change the wall follower's topic to /vesc/ackermann_cmd_mux/input/navigation. I fire up the joy node and I observe that I am only able to control the car with the joystick. When I switch over to the autonomous operation (by holding down right bumper RB button), I observe that the car is not controlled.

Then I change the topic to /vesc/high_level/ackermann_cmd_mux/input/nav_0 (based on sample solutions posted by MIT Racecar course TAs). Still nothing happens.

I want to make real time navigation with racecar. When I run the navigation packages, it sends me the data of type cmd_vel Twist. How can I check the motors using this data?

No commits in over a year. Is the MIT team still maintaining this project?

Based on the imu installation, how are the axis of the imu defined? I feel that the imu x direction is not aligned with the car forward direction. When I get the roll/pitch/yaw, which number tells me if the car is going up-hill or down-hill?

This device is very hard to find now. Any other VESC can replace this?

Hi,

Looking at the racecar/config/racecar-v2/vesc.yaml file I noticed the following lines:

# car wheelbase is about 25cm 
wheelbase: .25

which I believe is incorrect, the wheelbase is the distance between the centers of the front and rear wheels which is about 32cm in the case of a Traxxas 1/10 Rally Car. The official specs page confirms this.

Hi, is there any documents about how to collect training sample and train the car for a new course?

I built a Racecar J and could use teleop to drive it. I found, however, my car always turn right a little bit, even when I drive it straight. I think the steering is not centered. Is there anything I can tune with the BLDC tool? Or I have the tune the car mechanical parts?

I am wondering if there is any specification on sensors, car odom, etc that we can use to evaluate the status of the car and calibrate the car. I built a MIT racecar but when I ran gmapping in my garage which has simple features, the slam algorithm could not generate a good map. In some cases, It could not match lidar points of the wall after the car turned, even after bundle adjustment or loop closure, whatever it used. The same happened with google cartographer. So I believe it must be because of the car.

Can the car be controlled by orignal 2-channel radio system? If not, how could you drive the car to collect training video for a new course?

I have a question about the IMU message from MIT racecar package. The orientation quaternion from the message is related to the world system or the car base_link?

Currently, increasing the top speed of the robot requires changing the vesc.yaml, joy.yaml, and any other nodes that create throttle commands. Normalizing the speed controls would fix this.

According to the VESC config, the radian angles max out around +−0.264. Currently, clipping downstream in the VESC makes Gazebo sad. The clipping should move upstream in the node architecture.

Because stopping right now is really slow. Probably because the RPMs are hitting some max.

Alternatively, apply different gain on accel and deccel.

I have built a MIT-RACEcar and am using it for a fixed track race. I have my own particle filter which can localize the car. Once I know where I am, I can use pure pursuit to find the next target on the track. However, I do not know how to send out the ackermann control parameters. For example, I am currently at location (x1, y1), my near term target is (x2, y2). How can I compute the speed and steering angle from these two locations?

Since MIT-RACEcar has been used to race in the tunnel, it must have some code like this. Could anyone point me the location of the algorithm?

I used to use NiMH and everything works fine. To get more power, however, I replaced the NiMH battery with 4S LiPO. After that, I found that my car had difficulty to start. Please see the video here https://www.youtube.com/watch?v=D1XGa-vANRY I am not sure how to fix it and why I did not have the same problem with NiMH battery. Anyone uses LiPO battery and experienced the same problem?

I have a very strange problem I hope you can help me with. My MIT-Racecar can go forward and backward but it can not turn. Neither the teleop or ros command can make it turn. When I tried to turn the front wheels by hand, it feels very stiff but turntable. I am wondering what's going on and how can I debug. Is servo bad? Is there anyway to check with the BLDC tool?

I think there is a buy in the IMU node.

I used RVIZ to visualizae the /imu/data message. When I rotate the car orientation 90 degree, the axis ( or Yaw angle) in RVIZ rotated 360 degree.

Anyone found the same?

Why do you use lidar, stereo, rgb-d cameras? I'm curious for what reason you use lidar, stereo, rgb-d camera all of them.

Where did the ZED camera node from racecar-v1 go? Assuming this repo is maintained at all, does RACECAR still use a camera? Looks like both the ZED node and the structure.io camera are both missing from this codebase...

The hardware BOM includes a structure.io active stereo camera, but the actual rosinstall doesn't include any reference to a ROS node for controlling or publishing data from that camera. Is it no longer used on RACECAR?