b'107STEP 4 where the navigator calls out directions to theSTEP3driver, this constant dialogue between the two Road testing: Putting in thetest specialists ensures that what one sees inFleet release candidate process: hardest miles real life matches the digital depiction of theIntroducing new and improved vehicle sensors view on the others laptop. Thesoftwarea.Public road testing is perhaps the mostappearance of one or more cyclists on the road important aspect of Argos testing regimen.ahead of the vehicle, and their likely trajectory,a.Once the code has been thoroughly In fact, its the only way to validate that theprovides a clear example of the benefi ts andtested on a closed track, its bundled into Argo SDS works as needed. Exposing theimportance of commentary driving. a wider software release and enters whats SDS to everything that chaotic city traffic caffic canknown as the release candidate process. off er adds signifi cant value to the millionsd.Test specialists can intervene and offer adds signific Eff ectively quality control for new software Effectof simulations the company runs in virtualdisengage the vehicle from autonomousconfiguronfi gurations, the release candidate process testingand because every city is differy is diff erent,mode, either voluntarilythey are trained tois performed firormed fi rst in simulation, then at the Argo tests in multiple cities. exercise maximum cautionor because Argotest track, where the SDS is presented with Each test vehicle is monitored by twopolicy deems a takeover to be mandatorya mix of regular and less common scenarios, test specialists, one in the drivers seat, andduring certain phases of development, suchsuch as a proxy pedestrian appearing from one in the passenger seat. While out on theas when driving past a school, or if they seebetween parked cars or, in our passing a road, their task is to test specific prpecifi c processes,the flahe fl ashing lights of an emergency responsecyclist example, a bike up ahead swerving to such as a new pull-over feature for passengervehicle. The test specialist in the drivers seatavoid the sudden swing of a car door opening. pickups.is always ready to take control, by touchingThe release candidate process ensures the steering wheel or applying the brakes,the software release performs as it should, b.The team is asked to execute missions inand while they do so, their colleague logs thethat changes dont reduce overall system and around a geofenced area within the city,situation on the laptop. performance, and that it is acceptable and testing whether the vehicle found the correct location, pulled over, came to a completeSSTEPTEP safe for public road testing.stop, and unlocked the doors to accept a4 b.Any faults detected in the code result in the 4passenger.software being sent back to developmentand c.While operating the vehicle, the testso the cycle continues.specialists use the companys commentary driving technique. Inspired by rally car racing,4.d 4.d 4.c 4.c 4.b 4.b4.b4.a 4.a4.c 4.a4.d4.d 4.b 4.a4.cSTSEPSTTEPEP3.d 3.d3.d 3333.d3.c 3.c3.c3.c3.b 3.b3.b3.b3.a3.a 3.a3.aSTSSEPSTTTEPEPEP22222.a 2.b 2.c2.a 2.a 2.b 2.b 2.c 2.c2.a 2.b 2.cSTEP 2Closed-course testing: The top-secret test tracka.At the companys test track, Argo performsb.From simulation testing, the process movesc.If no faults are found, the software release structured testing, presenting the SDS withto live structured testing. Once the scenariothen undergoes initial public road testing with a wide variety of carefully staged challengeshas been constructed, the vehicle progressestest specialists in one or two test vehicles and scenarios, ranging from physicalalong the route in autonomous mode, and theon predetermined routes. Test specialists obstacles to so-called proxieshuman andproxy cyclist appears. When the engineers areevaluate the performance of the software animal mannequinsthat spring out in front ofsatisfied w release in a number of real-life scenarios, isfi ed with the systems performance near the vehicle. First, though, the test engineersproxies, they then begin testing with realincluding trafficaffic lights and unprotected left and perform in-vehicle simulations, where theycyclistsinitially, Argo test specialists on tworight turns.eff ectively trick the test vehicle into thinkingwheels instead of four.effectits facing a specific scpecifi c scenarioin this case,d.Once there are no issues with the that the car is following a bicycle. Likec.As with all testing, the goal is to find flao fi nd fl aws inperformance of the software release, and its virtual testing, in-vehicle simulation lacks thethe integration of new features. But althoughbeen approved for release, its rolled out to uncertainty of the real world, but it shows thesoftware configuronfi gurations can be changed atall vehicles in the Argo test fleeest fl eet to continue integration and test team how the SDS wouldthe track, new code can only be written at thepublic road testing in other cities. respond to any number of situations, such asdevelopment centers and must be reviewed passing a cyclist. before it can re-enter the test cycleFullBook_Mar24.indb 107 4/25/21 6:44 PM'