Need4Stek
The Need4Stek was a first year project at Epitech. It teaches you the basics of "artificial intelligence".
Autonomous car concept
Infos
| Info | Description |
|---|---|
| Module-ID | B2 - Introduction to Artificial Intelligence |
| Repo name | AIA_n4s_$ACADEMICYEAR |
| Binary name | ai |
| Language | C |
| Compilation | Via Makefile, including re, clean and fclean rules |
-
The totality of your source files, except all useless files (binary, temp files, obj files,...), must be included in your delivery.
-
All the bonus files (including a potential specific Makefile) should be in a directory named bonus.
-
Error messages have to be written on the error output, and the program should then exit with the 84 error code (0 if there is no error).
For this project, all of the lib functions are authorized.
By now you’ve understood that it’s now your turn to create your own autonomous car. It must be capable of driving on a track without hitting the walls or driving in the wrong direction. In order to help you with this task, several tools are avaible:
- CoppeliaSim is a simulator created by Coppelia Robotics. It enables the creation and control of robots. For this project, we have built a virtual track and car that you will use to implement for AI. You will find, further on in this document, how to install/use this software on your machine,
- an API (binary name: n4s). It is a communication interface that offers an array of actions that you can use (start a simulation, set the car motor’s speed, wheel angle,... ). You can also use it alone in order to test the communication protocol on the command line. This communication protocol is defined further on,
- a shell script (binary name: pipes.sh) that enables you to connect your program to the API and launch simulations. It’s a binary that you will run in order to test your AI. Its utilization is also presented further on.
CoppeliaSim
Installation
Download CoppeliaSim Edu V4.0.0 (the assets provided for the project function solely with Linux and Mac). In order to launch V-REP, you need to execute the onboard script:
Download and then unzip the latest version of the B-CPE-201_Need4stek_package.tgz from the intranet (file provided with the project description). Miscellaneous “.ttt” files are located in the “scene” file. These are the files that CoppeliaSim uses in order to save the scene simulations. As far as we’re concerned, they contain the track and the car to be driven. Before beginning your simulations, you must systematically open the track_1.ttt scene (or every other scene that you have made yourself) that is located in the CoppeliaSim (drag ’n drop or via the “File > Open scene... ” menu).
Make sure, when starting the CoppeliaSim, that the selected physics engine is the ODE.
The car will perform best in this mode. Menu: Simulation -> Using the ODE physics en-
gine
API
Overral Functionning
The n4s binary enables communication with CoppeliaSim (via a socket in C) and controls each of the ele- ments that we have put in our scene. It reads the commands sent to it on the standard input, carries out the task, and responds by writing on the standard output. Your binary should, therefore, do the opposite: give off its commands by writing on the standard output and receive the n4s responses by reading them on your standard input.
The command in the pipes.sh script enables you to correctly link the different binaries. You’ll need to execute the pipes.sh script in order to launch the simulation and see how it interacts with your AI.
Commands
The communication is done via a text-type protocol:
command value range answer type
START_SIMULATION - (1)
STOP_SIMULATION - (1)
CAR_FORWARD:float [0;1] (1)
CAR_BACKWARDS:float [0;1] (1)
WHEELS_DIR:float [-1;1] (1)
GET_INFO_LIDAR - (2)
GET_CURRENT_SPEED - (3)
GET_CURRENT_WHEELS - (3)
CYCLE_WAIT:int [0; INT_MAX] (3)
GET_CAR_SPEED_MAX - (3)
GET_CAR_SPEED_MIN - (3)
GET_INFO_SIMTIME - (4)
All the commands ends with\nand inevitably lead to an answer.
If n4s doesn’t receive a command, it won’t send out an answer.
Commands are not case sensitive.
For example, CAR_FORWARD takes a parameter that does indicate the engine power you want to drive
with (and not absolute speed): CAR_FORWARD:0.5\n Makes the car move forward at half of its maximum speed.
Responses
Here are the different answer formats, corresponding to each type:
| VALUE_ID | STATUS | CODE_STR | ADDITIONAL_INFO |
|---|---|---|---|
| (1): VALUE_ID | STATUS | CODE_STR | |
| (2): VALUE_ID | STATUS | CODE_STR | [:float]*32 |
| (3): VALUE_ID | STATUS | CODE_STR | float |
| (4): VALUE_ID | STATUS | CODE_STR | [long,long] |
VALUE_ID : Indicates the response code. The different values are listed below.
STATUS : Is whether “OK” or “KO”. It indicates if the command execution is a success or a failure.
CODE_STR : Corresponds to the verbal version of VALUE_ID. The different values are also listed below.
ADDITIONAL_INFO : Can contain information concerning the last checkpoint passed on the track (followed by the id of this checkpoint and the passage timestamp). The four different types of checkpoints are as follows:
Message Explanation:
First CP: Cleared first checkpoint passed
CP Cleared: Passing a checkpoint (other than the first and last on the track)
Lap Cleared: Complete lap
Track Cleared: End of the race. You need to stop the car and the simulation
VALUE_ID/CODE_STR correspondance:
| VALUE_ID | CODE_STR |
|---|---|
| 0 | Simulation has not been launched |
| 1 | No errors so far |
| 2 | Simulation running |
| 3 | Error. No details can be provided atm |
| 5 | Checkpoint error detected |
| 6 | Network operation failure |
| 7 | Server-side Error |
| 8 | Client-side Error |
| 9 | EOF reached |
| 10 | Simulation was correctly ended |
| 11 | Empty command |
| 12 | Unknown command |
| 13 | Wrong arguments provided. Please refer to protocol |
| 14 | Too many args provided with the command. Please refer to protocol |
| 15 | Pipeline failure |
| 16 | Unexpected command argument’s value |
| 17 | Camera infoget failure |
| 18 | Command not found |
| 19 | Simulation already up and running |
| 20 | CYCLE_DONE |
| 21 | Sensor reading failure |
| 22 | Scene must contain at least 3 CPs |
| 23 | Timer Init Failure |
| 24 | Timer get Failure |
| 25 | Failed to load Map |
FAQ
What is the viewing angle? 60 degrees, first-person viewpoint
What is the LIDAR unit of measurement? Millimeters.
What are the car’s dimensions? The car’s dimensions are not exactly known and can vary from one model to another (and very certainly between the virtual model and the real car). Try to have an AI that’s intelligent enough to function without this information. There are very simple ways to do it.
How much does the car weigh and what is its maximum speed? As with the previous question, this is information that you don’t necessarily need. What’s more, you just need to take measurements when the car starts up in order to deduce its speed.
How many drive-wheels does it have and at what angle is the car able to turn? Ditto.
Why don’t the values returned by the LIDAR exceed 3010? A real LIDAR uses a limited-reach laser. In order for the simulator to be as close to a real environment as possible, we have curbed the measured distances to this limit. In other words, when an obstacle is “de- tected” at a distance of 3010, it means that this obstacle is more than 10 feet from the car.
What do the two values in the type 4 response correspond to? They give, in seconds and nanoseconds, the time that has passed since START_SIMULATION.
What is the evaluation really looking for? An AI that is able to do laps, rather quickly, on the track (to give you a hint, tell yourself that a lap done in 5 minutes on a small track is much too slow). We’re also looking to see that your AI will stop driving and terminate the simulation in a dead-end. In this case, it must stop at least three feet from the wall, without touching it, before terminating the simulation.