Categories
News

RumiCar Simulator v12.1.0 released

We have released RumiCar Simulator v12.1.0. The release notes below are also available on GitHub.


v12.1.0

The shipped samples and the explanations now handle the values of RC_read as “distances” and “statuses” (Stage BR; the driving physics, the ranging, the learning API and the interpreters are untouched — driving with the shipped samples and the result of the shipped official sample race are unchanged. The three shipped language samples and the three Python car-type programs have the same verifyHash as v12.0.0 (= v11.1.0) on all 56 tabletop courses that have a finish. The official sample race’s verifyHash e9f1da0b and the frozen hashes f0–f3 are also the same as before, and only the resultSha256 of the result file was re-pinned because it includes the version (engineVer)). These are the same rules as in the real-car examples (the sketches of Arduino RumiCar-lib v0.7.1 and RasPi/Exercise-3.1 to 3.3 for the Raspberry Pi). Your program behaves as before — only the text of the shipped samples and the explanations changed. (1) The order of handling the values — first split a value of RC_read into a distance (0 or more) and a status (a negative number), and handle all statuses in one block at the start of loop(). On -1 (a sensor does not answer) stop with RC_drive(BRAKE, 0) and make no decision this time. -3 (out of range) is the status “nothing was measured before the limit in that direction”, and it is normal while driving. Treating it as “far” is a decision of the program. To compare it, replace it with the constant FAR_MM (9999) at the top of the program, and use FAR_MM only for comparisons (the left/right difference is calculated only when both sides are distances). (2) The limit is the constant LIMIT_MM at the top — write the value of the limit in the constant LIMIT_MM at the top of the program and call RC_setReadLimit(LIMIT_MM) in setup() (the samples of v12.0.0 wrote the number directly in the line in setup()). Keep distance thresholds at or below LIMIT_MM. (3) Output — the samples (C++, Python, JavaScript) now print the values as RC_read returned them, so -3 stays visible (the samples of v12.0.0 printed 9999, the value that replaced -3). (4) The simulator’s front, left and right sensors never return -1 — the real-car libraries return -1 on a sensor timeout. The simulator’s front, left and right sensors never return -1 (a dropout with sensor noise ON is also -3; this is unchanged from earlier versions). -1 is returned only by RC_read(REAR) on a car without the rear sensor. The “stop on -1” branch of the samples is there so that the same program can be run on a real car; a run in the simulator never enters it. The samples up to v12.0.0 put -1 into the comparisons as it was (on a real car a timeout would count as “close” and make the car reverse). (5) Explanations — the in-app explanations (Help chapter 7, Specs & Limits, the system guide, Physics Model, the car-type program notes) and §12 of the physics model document docs/physics_model.en.md follow this order, LIMIT_MM and FAR_MM. The physics model document now also states how C++ reads the arguments differently from the real car. The simulator’s C++ reads the arguments of the learning API by the same rules as Python: a fraction or true given as a direction is a bad direction (RC_read returns -2; RC_steer, RC_drive and RC_setReadLimit return 0 and do nothing), and a string of digits given as a strength, speed or limit is read as a number. A real Arduino converts a fraction or true to an integer first (0.5 becomes LEFT, true becomes CENTER). An ordinary program, which writes directions as constants and values as integers, shows no difference (the difference was found in the cross-check for v12.0.0; the interpreter has not been changed). (6) Samples whose style is unchanged — the C car-type programs (Apex Hunter etc.) keep the previous style using raw values and the confidence limit CONF. The nine full-scale competition programs (Circuit Racer, Range-Flow Estimator, Recon Racer and others) keep the previous style that replaces a negative value or a value beyond CONF with one large value (they are full-scale-only programs, not meant to be moved to a real car). (7) Verification — the verifyHash of the six samples (56 tabletop courses, production runRace) is the same as in v12.0.0. Run through the same flow as ▶, with sensor noise OFF and ON, the old and the new samples give the same car state (position, heading, steering, drive) from the start to the end of the run, and RC_read of the front, left and right sensors never returns -1 (six tabletop courses, 60 seconds each). The “stop on -1” branch was checked on the real-car libraries (RasPi/RumiCar.py 0.7.0 and RumiCar.cpp of RumiCar-lib v0.7.1) with inputs that include timeouts (with inputs without timeouts the commands equal those of the old samples on every pass; on a timeout the new samples output the brake only). Chapter D of the verification gate wf_bh6_docs feeds the interpreters a scripted RC_read and checks that a pass with -1 outputs only RC_drive(BRAKE, 0) and changes no state, that the commands of the three language samples match on every pass, and that the statements “the front, left and right sensors never return -1” and “how C++ reads the arguments” match the measurements. Exercise-3.3 and 4.2-Answer of RumiCar-lib v0.7.1 and Exercise-3.1 to 3.3 of RasPi (the current published versions) run through the interpreters unmodified (the texts in chapter D of wf_bo1_steer were replaced with the current versions). The permanent gates stay at 82, and browser/check_br1_values.mjs was added to the real-browser checks. Unchanged: the driving physics, the ranging itself, the learning API (the return values and warnings of RC_read, RC_setReadLimit and the others), the C++ / Python / JavaScript interpreters, the deterministic engine of official races (🏁), the saved-data format and the tabletop courses.


Verifying this release yourself

git clone https://github.com/RumiCar-group/RumiCar-Simulator.git
cd RumiCar-Simulator
node wf_run_all.mjs        # the full assertion-gate suite

The suite runs from a fresh clone with no dependencies beyond Node.js.
The real-browser checks under browser/ additionally need npm install and a display.

How this site is built (v1.6.0)

Some components are open source:

Enjoyed the machinery? The main event is our autonomous-driving simulator — it runs in your browser.
Try the simulator → / GitHub →