Mankind has been obsessed with and fascinated by autonomous vehicles for the last few decades. In the 1997 timeless classic – “Tomorrow Never Dies”, James Bond controlled and steered his car using his mobile phone and left audiences enthralled, wondering when they could do the same in their cars. Cut to 2020, and such features are a reality in some of today’s most advanced cars.

McKinsey & Company had predicted that driverless cars would be adopted by the masses by 2030 and few commercial self-driving cars or robotic taxis could be available as early as 2022.[1] So let’s take stock of where we are today and what the future looks like.

At the NextGen event in June 2019, BMW demonstrated level 4 capabilities in their flagship 7 series prototype, wherein passengers can summon the vehicle using a smartphone and also interact with the car using a tablet to request the car to go to a place of their liking! BMW has been testing L3 capabilities in the A9 autobahn for quite some time now and is anticipated to feature in the 2021 model of BMW iNext. [2] The University of Michigan introduced autonomous shuttles powered by Navya in their M-CITY facility. Though the shuttle travels at a top speed of 12 mph and can traverse only fixed routes, the introduction of fully autonomous vehicles in the public domain is itself a commendable achievement.[3] Waymo, also known formerly as Google’s self-driving car initiative, has even started offering autonomous robotic taxi services to customers in Phoenix, Arizona, without “safety drivers” in place. [4]

Browsing through such demos and events makes us feel like the future has arrived. However, the reality in today’s world is that such initiatives are only springing up in few pockets and geo-restricted spaces across cities. They’re actually far away in terms of mass acceptance and deployment of the technology. Lack of regulations, the massive cost of development coupled with expensive hardware systems, and limited consumer trust in travelling in a vehicle driven by a machine have all been hampering the future of autonomous cars. These have been the primary reasons for self-driving features not rolling out to production-ready cars just yet.

For most OEMs, the farthest they have been able to go in terms of production-ready cars is the Level 2 automation, which can also be classified as ADAS (Advanced driver assistance systems). Tesla’s autopilot, Polestar’s pilot assist, and Nissan’s pro pilot all classify as ADAS features, and have further ambitions to grow into a complete self-driving suite leading to L5 in the future.[5] So what obstacles are preventing them from making it to the finish line?

Automotive OEMs must find a way to ensure that full autonomous driving features get rolled out to consumers in a phased manner, given the lack of ecosystem to support such driving behaviors in most roads across the world. Over the last few months, however, manufacturers may have finally found the key to unlock potential through a simple yet critical aspect of the self-driving journey - the autonomous valet parking feature.

Autonomous valet parking is a mechanism by which cars can park themselves in designated or shared spots across parking lots once passengers disembark at a drop-off zone. This pioneering functionality uses all the key features and mechanisms that have been at the forefront of unmanned vehicle development.

From using sensor data to perceiving the surrounding environment, to autonomous compute platforms and the overlaying algorithms to steer the vehicle, plus the interaction with the surroundings (parking spots) using V2X technology, the vehicle performs complete autonomous driving albeit in a limited environment. This represents a victory for the industry, so much so that an increasing number of OEMs are now looking at it as a gateway to complete L5 driving.

Some of the typical features of the AVP suite include the existing ADAS functionalities like automated emergency breaking, automated parking assist and lane-keeping assist, which are coupled with autonomous driving features like parking slot identification, V2X interactions with parking sensors, and lane-change maneuvers.

An overview of Autonomous valet parking

Ford and Bosch have recently announced a partnership to demonstrate autonomous valet parking in Downtown Detroit. Systems in a test vehicle will communicate with sensors to locate an empty parking location and move the vehicle into the spot. This system includes safeguards that allow the vehicle to react and respond to objects and pedestrians in the drive path.[6] A similar partnership was announced between Bosch and Daimler, and the auto major received a nod from regulators to test the feature without a human safety driver. [7] Baidu has also announced a production-ready autonomous vehicle with the AVP (Auto Valet Parking) feature before 2021. [8] A surge in the pace of technological advancement has also seen valet parking robots being rolled out across Lyon airport.[9]

The increasing trend of investments in the valet parking feature proves that auto majors are betting big on it, as a step toward full autonomy on public roads. Let’s examine why:

1. The valet parking feature uses the exact same set comprising the sensor, compute and algorithm suites, while implementing and testing the capability in a limited environment of parking lots.
2. The total number of edge-case scenarios for parking lots is a lot lesser than public roads, where the interactions are with a lot more entities such as pedestrians, road blocks, repair works, traffic signs along with multiple types of vehicles.
3. The traffic density is much lesser in parking lots, therefore proving to be an easy test bed with less likelihood of interactions with pedestrians as well.
4. Once the feature is implemented and proved successful, it would lead to public acceptance of autonomous driving. This is evident through the success of the automated parking assist feature in many ADAS suites, which is now being replicated across newer vehicle models.
5. Easy to form a set of standards that could be implemented quickly across the globe along with standardized signs if the environment is limited. The parking lot is a perfect environment to start with.

While the developments have been exciting and technology has been instrumental in positioning autonomous driving as the future of mobility, for auto OEMs to make this a standard in production-ready vehicles, such functionalities must also end up being cost-effective. Else, it will not make sense to roll out such expensive features to the public.

One of the major challenges has been the overreliance on sensor suites across parking lots to power the AVP feature. This makes the solution costly and needs all garages to be transformed with adequate infrastructure before AVP can be rolled out. Hence, it is imperative that the vehicle becomes smarter through the use of intelligent technology that can detect parking spaces and empty parking lots by itself.

There has been a focus across the globe on improving capabilities from the vehicle’s end and making business cases for such technology through aggressive technology advancement and getting rid of overreliance on expensive suites, while making vehicles safer. Clearly, there is a long road ahead before it can become a standardized feature across vehicles.

Wipro autonomous systems and robotics has been pioneering the development of the autonomous valet parking feature. It is a part of Wipro’s AI Algostack suite for autonomous driving. The AVP feature here uses only RADARs and Camera as primary sensors to navigate the vehicle, making the solution cost-effective and safe. The feature has been tested using the SDV (Self-driving vehicle) in a box simulator for autonomous vehicles along with multiple use-cases such as parallel parking, perpendicular parking, and angular parking, which can either be in open parking lots in neighborhoods or closed parking lots such as in apartments, office and shopping complexes. The vehicle uses HD maps associated with each parking lot to localize and navigate around the surrounding environments. The solution has also been tested against multiple edge cases to make it robust and fail-safe.

Wipro’s Autonomous Valet Parking suite under simulation using SDV-in-a-Box simulator

Wipro’s autonomous valet parking is part of its modular autonomous driving software suite (also known as AI Algostack) with sensor and driver behavioral data analyzed on the vehicle (edge), and key info and learnings channeled back to the data lake, which can be on cloud or on-premises. One of the ground-breaking features includes the use of RADAR instead of LIDAR for complete autonomous valet parking, which makes the solution highly optimized.

Wipro is committed to envisage a cleaner and greener planet for everybody, using visionary and innovative solutions to realize the dream of self-driving cars. Visit us here to know more.

References:

1. Automotive revolution – perspective towards 2030, McKinsey & Company, January 2016
2. BMW Demonstrates Level 4 Autonomous Driving In the 7 Series; https://www.autoevolution.com/news/bmw-demonstrates-level-4-autonomous-technology-in-the-7-series-135555.html; June, 2019
3. Driverless Shuttles Debut at University of Michigan; https://www.govtech.com/fs/automation/Driverless-Shuttles-Debut-at-University-of-Michigan.html; June 2018
4. Waymo begins offering robot-taxi trips sans safety driver in Phoenix; https://venturebeat.com/2019/10/28/waymo-begins-offering-robot-taxi-trips-sans-safety-driver-in-phoenix/; October 2019
5. ADAS tech remains a blind spot for independent car sellers — and it’s putting lives at risk; https://thenextweb.com/shift/2020/08/13/adas-tech-remains-a-blind-spot-for-independent-car-sellers-and-its-putting-lives-at-risk/
6. Ford, Bosch and Bedrock announce an automated valet parking garage in Detroit; https://techcrunch.com/2020/08/26/ford-bosch-and-bedrock-announce-an-automated-valet-parking-garage-in-detroit/; August 2020
7. Daimler and Bosch’s driverless parking gets OK to operate without human supervision; https://techcrunch.com/2019/07/23/daimler-and-boschs-driverless-parking-feature-can-legally-operate-without-human-supervision/; July 2019
8. Baidu self driving cars will debut before 2021 for autonomous valet parking;   https://www.drewreportsnews.com/2020/08/02/baidu-self-driving-cars-will-debut-before-2021-for-autonomous-valet-parking/; August 2020
9. Stanley Robotics, Lyon Airport to expand robotic parking service; https://www.roboticsbusinessreview.com/regional/stanley-robotics-lyon-airport-to-expand-robotic-parking-service/

Subham Chatterjee

Subham works to power the business development and GTM strategy for Wipro autonomous systems and robotics. He also leads product management for SDV in a box – a global scale simulator for autonomous vehicles. Subham specializes in delivering business outcome driven solutions on simulation, validation, and testing for autonomous mobility. He holds an MBA from IIM Raipur and is passionate about the world of possibilities using autonomous mobility. He can be reached at subham.chatterjee1@wipro.com.

Natrajan Vittal

Natrajan is a part of Wipro’s AS&R (Autonomous Systems & Robotics) practice, which has the distinction of having developed an in-house and complete Autonomous Systems & Robotics product suite, including its flagship Autonomous Shuttle – the “Wipod”. His contributions have been supporting the product and business development. He holds a Masters in Engineering degree from IISc Bangalore. He is passionate about technology making a difference in people’s lives, and can be reached at natrajan.vittal@wipro.com.