Given the advent of autonomous cars and their growing penetration in the transportation market, it is empirical that companies consider using Swarm Intelligence to tackle traffic and make the daily commute a lot more pleasant for billions around the world. Currently autonomous cars act as individual units. Swarm Intelligence focuses on communicating and learning collectively instead of acting as an individual unit. Benefits of Swarm Intelligence in Autonomous Cars Maintain a constant average speed with other swarm units and ensure a smooth flow of vehicles in the swarm Warn and redirect other nearby vehicles if bad road conditions ahead (black ice, wet/slippery conditions, potholes, barrier in the middle of the road) Warn and redirect others if road/traffic accident ahead. Warn and redirect others if there is an upcoming heavily congested lane Warn and redirect others if a vehicle is seen driving rogue. Higher safety as a swarm since the workload of collision detection and navigation is done collectively Communicate other swarm vehicles of a parking spot that was just vacated Communicate with other swarm vehicles and prioritizing vehicles for lane merging and changing Cooperation with other vehicles in the swarm that will lead to a smoother flow of traffic Common algorithms used in Swarm intelligence - Density of virtual pheromones deployed by previous units in swarm affects which path other units in the swarm take. The swarm unit who reaches the destination first is the unit taking the shortest path. The highest density is found in the shortest path since more swarm units would take this path. Swarm units coordinate their path socially and follow the trail of pheromones. Communication and coordinates happen through the pheromone - like in case of ants. Ant Colony Optimization - Similar to birds flocking on to a target and chirping louder to alert others as they are closing in. Swarm unit closest to target tweaks its value. Each iteration of change in those values results in a better solution. In the end, the tweaked value of other swarm units will be close to the value of the swarm unit which is closest to the target. Particle Swarm Optimization - Individual swarm units figure out what positions to be in and then move from regions of high population to low within a swarm to even out the flow. Shape formation Algorithm in a swarm - Comprises of standard waypoint following, avoid (move around), exchange (wait for the other swarm unit to leave in case of head-on collision), go through (wait until the other unit in the swarm has passed at an intersection), standard docking, wait until prior units finish docking, standard safe distance from other units Traffic Collision Avoidance in a swarm - Sub-swarms within a swarm are assigned part of the workload depending on the current situation of the environment. Constitutes the Card Dealer’s Algorithm and Extreme Communication Algorithm. Dynamic Task Assignment Algorithm Swarm Algorithms in action Below are examples of 3 algorithms in action that allow individual swarm units to act collectively and cooperatively as a swarm to avoid obstacles, flow easily through choke points and regroup back in free space! Figure: Trajectories of swarm intelligent units in a roundabout and chokepoint traffic scenario as seen in many cities across the world. imagesrc: Boeing and UWashington researchers Controlling the swarm Further, there are many models on how a swarm can be controlled. Latest Developments Centralization vs Decentralization Most swarms tend to have a lead unit/vehicle and decisions are centralized. The success or failure of the swarm depends on one single leader in this case. In cases of swarms with a lead, the chances of the whole swarm failing to continue operations is high. A new algorithm released days ago by researchers at Northwestern focuses on decentralization. This new advance promises collision-free and deadlock-free movement of a swarm. It has been tested on a swam of 100 real units (1024 simulated) that were able to outperform other approaches. This decentralization ensures that even if one unit fails in a swarm, the other units continue their operation as a swarm. This is advance is critical given any given sensor in a vehicle/individual unit can fail at any given time. In the decentralized model, an individual unit travels within a grid and communicates only with other units in swarm close by to figure out the best position of it in the swarm. The individual unit doesn’t change position until the other position in the grid is vacant. What the future holds Instead of a single car computing its own route to the destination, the car will be joining multiple swarms throughout its journey to the final destination to ensure a smooth flow of traffic. Smart Cities So far cities have only explored building more highways, wider lanes, or bridges to improve on traffic conditions. All of the above are limited in length or width due to existing infrastructure on the surface such as existing buildings, parks, etc. We aren’t limited by such things if we go underground and build tunnels. A whole labyrinth of underground tunnels can be created in a given city to streamline traffic. A city can build dozens of layers of such underground tunnels. These IoT tunnels can be equipped with cameras at strategic points that can count each car coming in and current cars in any given tunnel and redirect coming traffic to a tunnel with less number of vehicles. Cameras inside the tunnel can count the number of cars in any given tunnel. Upcoming traffic can be directed to a tunnel with less number of vehicles. This can be done by communicating with the autonomous car using or by displaying a changing digital sign that says ‘tunnel closed’ or ‘tunnel open’ at the entry of each tunnel. vehicle-to-infrastructure (V2I) communication V2V and V2I communication systems The other end of such underground tunnels can have dozens of openings across strategic points in the city. What the future should and hopefully will look like: The cities of the future will be mostly be reserved for pedestrians apart from a few areas for vehicles. Further, the city views and the city environment will be much more pleasant and pedestrian-friendly than present day.

Flow

Boeing

Target

How Intelligent Systems and Robots Will Be Designed In The Near Future

How AI Could Save the 3D Printing Industry and the Future of Machines

Follow me for more on state of the art in machine learning & AI

Read My Stories

Too Long; Didn't Read

Make resilience your competitive advantage

How Swarm Intelligence-Powered Vehicles and Smart Cities can Help Streamline Traffic

About Author

Comments

TOPICS

THIS ARTICLE WAS FEATURED IN

Related Stories

Untitled Story

Introducing CatalyzeX: A Browser Extension for Machine Learning

24 Stories To Learn About Automotive

31 Stories To Learn About Autonomous Cars

4 Overlooked Challenges to Our Shiny Self-Driving Future

46 Stories To Learn About Self Driving Cars

Introducing CatalyzeX: A Browser Extension for Machine Learning

24 Stories To Learn About Automotive

31 Stories To Learn About Autonomous Cars

4 Overlooked Challenges to Our Shiny Self-Driving Future

46 Stories To Learn About Self Driving Cars

Light-Mode

Classic

Newspaper

Dark-Mode

Neon Noir

Minty

HN StartUps