b'82For experienced drivers,getting from point A to B in a city might seem simple. They survey their surroundings, pick a route, and punch the accelerator. But this process only appears simple because humans are naturally gifted at quickly inferring how others around them are likely to act, so that they can predict their next moves and react appropriately. A good driver will observe another car approaching a cyclist in the distance and decide to slow down to allow that other car to merge into their lane to avoid colliding with the cyclist sharing its lane. Veteran drivers are also good at driving socially and defensively. Socially means they drive in a way that clearly communicates their inten- SENSINGtions to everyone else on the road. For example, a driver knows to inchBefore it venturesonto public streets, a self-driv-forward at a stop sign to signal that theyre about to proceed. Drivinging vehicle needs to have an in-depth understanding of its defensively means they anticipate problems before they occur and takesurroundings, and an ability to predict the future actions actions preemptively. A defensive driver might slow down to increaseof all the other actors on the roads. For Argo, this starts their following distance if a car in front is acting erratically. This kind ofwith two things: the SDSs comprehensive 3D map of its behavior demonstrates situational reasoning that is natural to others onenvironment, and the vehicles perception system.the road. Unusual behavior, on the other hand, can look unnaturalsay,Argos 3D maps capture the entire spectrum of static a driver maintaining an overly-large following distance even though sur- streetscape objects and traffic flow dynamics, equipping rounding traffic is flowing normally, or a driver ignoring apparent risksArgo self-driving vehicles with prior knowledge of a by tailgating closely behind that erratic car. city. The maps dont just demarcate the intersections, av-The combination of safe, defensive, and socially appropriate driv- enues, and side-streets that make up a dense urban envi-ing is what we call naturalistic driving. And while developing a saferonment. They also include traffic signal locations; street self-driving vehicle is already a challenge, achieving the level of reason- signs; local laws and regulations, including speed limits; ing and control to achieve naturalistic driving is far more challenging. Itcrosswalksandlanegeometries;bikelanesandcurbs; takes an army of software developers, machine-learning and computerbuildings and landmarks like statues; even trees.vision specialists, test engineers and specialists, data analysts, map an- Using this wealth of contextual knowledge, an Argo notators and object labelers, and more. For self-driving technology com- self-driving vehicle can then compare what its observing pany Argo AI, it means designing its self-driving system (SDS) to sensein real-time with the detailed 3D model of the physical and understand the vehicles surroundings, plan future actions, and actspace. Having a complete understanding of its surround-on that plan safely and naturalistically.ings enhances the Argo vehicles ability to move safely Here, lets unpack all the steps that go into making naturalisticand naturalistically through the world. driving possible.Argos perception system fuses data from on-board sensorsmade up of cameras (stereo and monocular), radars, and lidar (short and long range)to perceive ev-ery actor sharing the road as well as any street markings that may impact driving conditions. The Argo detection team takes that perception data and uses it to train deep learning models that recognize everything in the vehicles environmentincluding other vehicles, pedestrians, cyclists, animals, road signs, traf-fic lights, sidewalks, vegetation, and construction zones. Simpler self-driving systems do not recognize all these objects, which limits their ability to interact more nat-urally with others sharing the road. For example, some systems might only recognize traffic barrels, rather than a construction zone, and therefore might cut in between barrels to get back into a lane that the barrels were at-tempting to mark as closed.The detection team supplies image-based information to the Argo tracking team, which fuses it with sensor data to model the movements of the classified objects. This helps the SDS construct a complete, dynamic picture of everything in its environment.FullBook_Mar24.indb 82 4/25/21 6:43 PM'