Self Driving Cars

An Introduction to

@aktech

The First Question

Why?

1885

Later that Year

First Public Test Drive

It crashed into the Wall!

130 Year Since then

We have been working around the least reliable part of the Car

The Driver!

We have

Made the Car stronger

Seat Belts

Air bags

ABS

Bla Bla .. Bla ...

Reality is

1.25 million deaths worldwide

We need to Fix the Bug rather than patching around it.

The Driver!

Here Comes into Play

Self Driving Cars

Motivation

• to eliminate road deaths caused by human error,
• reduce traffic, and
• free up time spent commuting

Technology

Lidar

Light Detection & Ranging

• creates 3D images of objects
• calculates how far an object is from the moving vehicle
• based on the time it takes for the laser beams to hit the object and come back.

Lidar

Front Camera For Near Vision

helping the car ‘see’ objects right in front of it.

• These include the usual suspects - pedestrians, and other motorists.

• ​also detects and records information about road signs and traffic lights

Bumper Mounted Radar

4 front & Back

Bumper Mounted Radar

• aware of vehicles in front of it and behind it.
• software is programmed to (at all times) maintain a distance of 2-4 seconds (it could even be higher) vis-a-vis the car ahead of it.
• the car will automatically speed up or slow down depending on the behaviour of the car/driver ahead.
• keep passengers and other motorists safe by avoiding bumps and crashes.

Aerial that reads Precise Geo Location

Aerial that reads Precise Geo Location

• receives information about the precise location of the car,
• GPS estimates may be off by several metres due to signal disturbances and other interferences from the atmosphere. To minimise the degree of uncertainty, the GPS data is compared with sensor map data previously collected from the same location.
• As the vehicle moves, the vehicle’s internal map is updated with new positional information displayed by the sensors.

Ultrasonic sensors on Rear Wheels

Ultrasonic sensors on Rear Wheels

• helps keep track of the movements of the car and will alert the car about the obstacles in the rear.
• already in action in some of the technologically advanced cars of today. (Cars that offer automatic ‘Reverse Park Assist’ )

Synergistic combining Of Sensors

Synergistic combining Of Sensors

• All the data gathered by these sensors is collated and interpreted together by the car’s CPU or in built software system to create a safe driving experience.

Programmed To Interpret Common Road Signs

• At the moment, before a self-driven car is tested, a regular car is driven along the route and maps out the route and it’s road conditions including poles, road markers, road signs and more.
• This map is fed into the car’s software helping the car identify what is a regular part of the road.
•  As the car moves, its Velodyne laser range finder kicks in and generates a detailed 3D map of the environment at that moment.
• The car compares this map with the pre-existing map to figure out the non-standard aspects in the road, rightly identifying them as pedestrians and/or other motorists, thus avoiding them.

Breaking Down the Technicals

• Mapping and Localization
• Obstacle Avoidance
• Path Planning

Mapping and Localization

• laser rangefinder scans the environment using swaths of laser beams and calculates the distance to nearby objects
• video from camera is ideal for extracting scene color
• The vehicle filters and discretizes data collected from each sensor and often aggregates the information to create a comprehensive map, which can then be used for path planning.

• It uses its GPS, inertial navigation unit, and sensors to precisely localize itself
•  localization algorithms will often incorporate map or sensor data previously collected from the same location to reduce uncertainty
• As the vehicle moves, new positional information and sensor data are used to update the vehicle’s internal map.

Obstacle Avoidance

• Obstacles are categorized depending on how well they match up with a library of pre-determined shape and motion descriptors.
• uses a probabilistic model to track the predicted future path of moving objects based on its shape and prior trajectory.

For example

• If a two-wheeled object is traveling at 40 mph versus 10 mph

It is most likely a motorcycle and not a bicycle and will get categorized as such by the vehicle.

Path Planning

• The goal is to use the information captured in the vehicle’s map to safely direct the vehicle to its destination while avoiding obstacles and following the rules of the road.

• manufacturers’ planning algorithms will be different based on their navigation objectives and sensors used

Path Planning

In general

• The algorithm determines a rough long-range plan for the vehicle to follow while continuously refining a short-range plan

(e.g. change lanes, drive forward 10m, turn right).

Path Planning

A general Algorithm

• It starts from a set of short-range paths that the vehicle would be dynamically capable of completing given its speed, direction and angular position

and removes all those that would either cross an obstacle or come too close to the predicted path of a moving one.

Path Planning

• For example, a vehicle traveling at 50 mph would not be able to safely complete a right turn 5 meters ahead,

therefore that path would be eliminated from the feasible set.

Path Planning

Remaining paths are evaluated based on

• safety,
• speed, and
• any time requirements.

Path Planning

Once the best path has been identified, a set of

• throttle,
• brake and
• steering commands

are passed on to the vehicle’s on-board processors and actuators.

Path Planning

takes on average 50ms

Path Planning

It can be longer or shorter depending on the

• amount of collected data,
• available processing power, and
• complexity of the path planning algorithm.

The process of localization, mapping, obstacle detection, and path planning is repeated until the vehicle reaches its destination.

The Road Blocks

• GPS can be unreliable
• computer vision systems have limitations to understanding road scenes
• variable weather conditions can adversely affect the ability of on-board processors to adequately identify or track moving objects
• roads without clear lane markings (Difficult to change lane)
• Driving in cities is much harder for autonomous cars than cruising on the highway: (Sloppy GPS (tall buildings), Easy to miss something)

References

• http://robohub.org/how-do-self-driving-cars-work/
• Chris Urmson: How a driverless car sees the road | TED Talk
• https://www.national.co.uk/tech-powers-google-car/

Thanks!