Human versus Computer Performance - 2014 Automated Vehicle Symposium SF Serge Lambermont
2014 Automated Vehicle Symposium - Serge Lambermont
The Tuesday breakout session discussed the concept of digital infrastructure and mapping in order to begin
exploring the following questions:
What elements are included in the definition of digital infrastructure?
What data are needed and how will the data used by the vehicle? By the infrastructure?
What is the government’s role and responsibility in digital infrastructure and mapping?
What is the relationship between industry and infrastructure owners and operators in an automated vehicle environment?
Mr. Serge Lambermont’s presentation provided an overview of digital infrastructure
and other interrelated activities supporting the development of automated vehicles,
including the role that suppliers play in this environment. Two key areas were
highlighted: 1) understanding the strengths and weaknesses of technology today
and how digital infrastructure can be used to further support automated vehicles;
and 2) exploring the uses and impacts of digital infrastructure in adjacent markets
today that could then lead to further developments in automated vehicles. The
following are points raised from these two key areas:
Human versus computer vision/lidar/radar detection and classification:
Assuming that human are perfect drivers, how would this compare to machines? A quadrant diagram with computer recognition and human recognition on each axis was presented to demonstrate the strengths and weaknesses of both groups (see figure below)
Maps for localization and lane markings can immediately address areas where human recognition is strong but machine recognition is lacking, including road boundary detection, curved lane marking, crossing lane marking, adjacent lane marking, stop lines and lines for lane marking.
ADAS maps can address stop lines, traffic lights, traffic signs, and lines for lane markings, which can be recognized fairly consistently by machine but this could provide improvement and redundancy.
V2I functionality can be used for determining trajectories of other vehicles/objects, and reading traffic lights and traffic signs, as well as V2V redundancy.
Adjacent market developments and impacts on Digital Infrastructure
figure, human versus computer detection and classification
Passenger Car Market & Commercial Vehicles: This market continues to
be driven by improvements to active safety and New Car Assessment
Program changes. This includes automated braking, detection systems,
active collision avoidance, automated highway lane change. This market
uses digital infrastructure for ADAS maps, V2X, functional safety and
Low Speed Autonomous – no steering wheel: This is a somewhat new
market; it is envisioned that these vehicle types would primarily be used
for retirement, airports, entertainment, work campus, care free city
centers, manufacturing, etc. There is more confidence today for
programs at low speeds; we are seeing a large number of low speed
prototyping vehicles emerging. These vehicles use maps but are also very
dependent on the sensor suite. Digital infrastructure is used for
Tech enabled car share and rideshare: This market is already using
digital infrastructure to a small extent as information is being transferred
to the cloud and the vehicle. This requires dynamic digital infrastructure
– optimal booking-routing, real-time cloud based updates, new physical
infrastructure, and call on demand.
Personal Rapid Transit: This includes smaller vehicles (2/3 wheel, limited
width), coming to market for fuel efficiency and individual mobility
purposes. These vehicles may need to share lanes with regular vehicles
and could impact the use of digital infrastructure.