Intelligent Vehicle Systems

Transcription

1 Intelligent Vehicle Systems Southwest Research Institute Public Agency Roles for a Successful Autonomous Vehicle Deployment Amit Misra Manager R&D Transportation Management Systems 1

2 Motivation for This Presentation Motivation: o Traffic deaths have plateaued in the US at around 32K a year: o Legislation is NOT working to reduce deaths Question: Can technology provide a solution? YES o Connected Vehicle (CV) o Automated Vehicle (AV) This presentation will explore: o Current state-of-the-practice o What can public sector do to get ready for CV and AV (CAV) 2

3 Future of Connected Vehicle. Technology has been researched and is ready to deploy V2V is easier because it does not require infrastructure clear safety benefits (justifies the cost) V2I or I2V is challenging because of the infrastructure (and continuing maintenance) costs More and more real world demonstrations occurring 3

4 Planning and Preparation for CV Deployment Identify and prioritize your needs Learn: o Understand the potential applications that can help address those needs o Understand the system architecture and how it relates to your existing system architecture o Understand what has been done to date and what is being done elsewhere Realizing CV: o Start small and iterate to expand the deployment area and capabilities o Measure performance and evaluate during each iteration to inform further deployment 4

5 Short Summary of AV Technology State of Technology: o Semi-Autonomous: Available today o Full Autonomy: Not yet Connected Autonomy: A likely reality Short Term: Adopting connected vehicles (V2V and V2X) is preparing for autonomous vehicles 5

6 NHTSA / SAE Driving Levels Partially Automated Driving exists today Source: SAE Autonomy limited to specific driving environments Requires human fallback Semi-Autonomous Driving available TODAY SAE and NHTSA levels different 6

7 Automated Vehicle Technology Basic question: o What is the PURPOSE of a driverless vehicle? Possible answers: o Ultimate solution to the driver distraction problem o Should reduce accidents (although until a significant penetration the overall effect is questionable) o Should enable a reduction in traffic fatalities o Make transportation systems much more efficient (more vehicles in the same space) 7

8 How will mapping data be handled? New Road Construction Almost 14,000 miles of new roads built annually in the US 2016 Copyright SwRI 8

9 How will mapping data be handled? It is clear that the industry needs a new kind of intelligent sensor a "live map" that provides the vehicle with an awareness of the road environment beyond the reach of its other on-board sensors. HERE Rumor: Google s reason for advancing AV technology Source: HERE Roadway Data: Delivered in real-time Centimeter lane level accuracy GPS, photo, and point cloud Petabytes of data 2016 Copyright SwRI 9

10 How will mapping be handled? Challenges Real-time updates to: Changes in roadway infrastructure Road closures Conditions Construction lane changes Distributing large data set on a national scale in real-time HERE financially backed by Audi, BMW, and Daimler Source: HERE Source: Google 2016 Copyright SwRI 10

11 How will mapping be handled? Your Data Will Have Value Traffic data for public: Nice to have Roadway map data for AVs: Must have New roadways require pre-mapping Capturing and real-time distribution of map data: Complex and expensive Commercial sector moving Conclusion: Commercial sector managed Actions to Consider: Short-term: Build relationships and partnerships Short-term: Commoditize or find a value proposition Medium-term: Plan to allow mapping providers advanced access to new roadways 2016 Copyright SwRI 11

12 How will Traffic Operations Change? Negative Obstacle Avoidance Challenging for AV s to detect Low-probability, high-impact problem Technical need for a high fidelity observed world model that is dynamically updated in real-time Copyright SwRI 12

13 How will Traffic Operations Change? Systems need to be capable of adapting to dynamic traffic patterns, construction/work zones, accidents, weather, etc Copyright SwRI 13

14 How will traffic operations change? Proposed real-time environment data : Construction Traffic congestion Lane closures Accidents Weather-related changes Variable traffic regulations Who is the best source of this kind of data? You are! (Road Operators) Monitor for events Verify events Know when events clear 2016 Copyright SwRI 14

15 How will traffic operations change? Autonomy Sensors dual-purposed Detect accidents Report traffic conditions Find potholes Traffic operations information sources Current: DOT managed 911, DOT sensor networks Future: Vehicles as probes Auto OEMs, Google (Waze), etc CV Infrastructure (V2X) Actions to Consider (All short-term): Be ready for more trends like Waze Research ways to communicate construction Use analytics to parse big data 2016 Copyright SwRI 15

16 What changes need to be made to the roadway infrastructure? Building our way out of congestion: Does this problem go away? Obtain 3,000, 4,000, or more vehicles/hour/lane? Some say even more and some say no Can we narrow lanes? Reduced accidents translates to less capacity to handle nonrecurring congestion Transition period: mixed autonomy and human driven vehicles Efficiencies will be hard to gain Technology Lanes Next evolution to HOV and express-lanes Actions to Consider: Medium-term: Planning to facilitate technology lanes Long-term: Planning requires a full understanding of autonomous vehicle throughput / density Research of autonomous vehicle throughput / density needs further funding 2016 Copyright SwRI 16

17 What changes need to be made to the roadway infrastructure? What about roadway signage? Expensive Perception in poor conditions (weather) Visually occluded Knocked over Actions to Consider: Short-term: Dynamic content: Adopt Connected Vehicles travel advisory messages (TAMs) Static content: Likely handled by mapping firms. Will DOT deploy virtual signs? Long-term: maybe no physical signs 2016 Copyright SwRI 17

18 What changes need to be made to the roadway infrastructure? Does lane stripping, centerline markers, and other road markings matter? Now: Yes! Future: Probably not Actions: Short-term: Road markings are important. Long-term: a future of no markings (or barriers?) 2016 Copyright SwRI 18

19 What are the security implications? Automobile Security 200+ electronic control units 100M lines of code Multiple suppliers Cars are complex Attack Surfaces of AV/CV Environment Vehicle Wireless communication DOT / Road Operator infrastructure Actions to Consider: Short-term: Build a culture of cybersecurity into your agencies. Treat it like Safety. Short-term: Take steps to secure your ITS infrastructure 2016 Copyright SwRI 19

20 2016: Autopilot How soon is full automation? o New models from Tesla and Cadillac can cruise at highway speeds with no driver input. 2020: Some what self-driving car o Low-speed, partially autonomous vehi`cles may be permitted in controlled settings like retirement communities. Also possible: dedicated highway lanes for self-driving cars. 2025: Mostly self-driving car o Highly automated driving will be allowed on more roads, but drivers will still need to be able to take over in certain situations. Public transportation could well be driverless. 2030: Fully self-driving car o People will be able to summon driverless cars anytime to take them anywhere. Long-haul cargo delivery will be autonomous. Embrace the technology and prepare as much as possible! 20

21 Intelligent Vehicle Systems Southwest Research Institute Thank You Amit Misra