Light rail drivers will inevitably face challenges — a car suddenly brakes in front, an animal crosses by, or a bicycle runs on the tracks. In addition, distractions can pull the attention away from the driver, and the driver and all passengers may suddenly face a potential collision. Drivers need to react quickly and decide whether to continue or stop.
And if a collision happens, are drivers fully responsible? And are some accidents unavoidable? Is there any assistance to help reduce the risk of collisions?
A traditional collision detection and warning system installed in light rails might help avoid collisions to some degree. These systems use a sensor to detect nearby obstacles and would get the driver’s attention with audio alerts (bells and rings), but they won’t interfere with the controls of the light rail. Advancement to this alerting system is video data and automation.
An intelligent video system should be capable of automatically helping avoid a significant collision to a greater degree. However, this system will not work where there are poor light conditions, such as in underground or tunnel railways or when the weather impedes full visibility.
Other challenges with a light rail automation system :
- Lack of full connection coverage. Automation systems may rely on Internet connectivity or connection to a high-end server for added intelligence. Connectivity will fail in wireless dead spots.
- Dependency on cloud computing. A light rail system may be dependent on cloud computing for its automation or intelligence.
- Costs associated with bandwidth. Moving raw data from the light rail to the enterprise data center or the Internet can be costly in terms of bandwidth.
- Demand for ultra-low latencies. The need to react quickly to avoid collisions demands ultra-low latency networks when the processing power is on a remote high end-server.
A Machine Vision (MV) system for light rail vehicles will address the abovementioned challenges. It helps complex tasks like real-time collision avoidance reliably and consistently. This automatic collision warning system is capable of detecting and classifying objects in different lighting and weather conditions. The system detects cars, buses, nearby rail vehicles, and static and abnormal track obstacles.
If the solution detects any sign of danger (within a predefined area), it will generate a real-time alert and trigger an automatic response, such as pulling the brakes.
This MV system is based on the following:
- Data collection devices such as video, radar, or electro-optic sensors and a combination of AI and MV technologies running on the edge.
- An onboard processing device— an edge computing device capable of processing AI and MV models (Edge AI) right on the light rail.
Why Edge AI?
Traditionally, AI models have only been possible via robust data centers usually located in the cloud. Edge, on the other hand, is about processing data as close as possible to its source, unlike cloud computing, which centralizes computing on a remote data center. Now, sophisticated edge computing devices may also be capable of processing Artificial Intelligence AI— thus Edge AI.
Edge AI combines edge computing with AI. Machine Vision uses AI to give computers meaningful insights from visual inputs (photos and videos). In Edge Computing, collected data is processed as close as possible to its source or at the network’s periphery. With edge AI, organizations and businesses can start getting value from the large amounts of collected operational and sensor data. Edge AI also enables automation and improves the safety and efficiency of applications— benefits that can significantly impact use cases like light rail companies.
Edge AI Appliance: EAI-I130.
Lanner’s EAI-I130 is an industrial-grade edge AI inference system running NVIDIA® Jetson Xavier NX/Jetson Nano. This appliance supports up to 21 TOPS (Trillions or Tera Operations per Second) for AI performance, which is an excellent fit for compute-intensive workloads. In addition, the EAI-I130 appliance also supports 5G and Wi-Fi6 simultaneously to improve the range of communications. Having advanced wireless support guarantees low latencies, better bandwidth, and coverage whenever needed.
This industrial-grade appliance is also built to withstand extreme environmental conditions. It is created using the IP40 standard with a fanless design capable of operating within the -40°C To 70°C temperature range. Ingress Protection level 40 (IP40) is protection from solids, such as tools and small wires greater than 1 millimeter.
Other EAI-I130’s key features:
- 2x GigE PoE LANs with support for IEEE 802.3 af/at PoE(+). Great to power up video cameras with PoE.
- 1x HDMI. To connect a monitor and keep track of any data from inside the rail.
- Rich I/O for added capabilities. I/O including 2x USB 2.0, 2x RS232/422/485 (COM1 & COM2) and 1x CAN 2.0A/B (Optional For COM2).
Below is a brief list of the possible benefits of using the Machine Vision solution for light rail collision avoidance.
a. Detect potential accidents. This MV system based on edge AI can detect an object, determine the distance to it, calculate the probability of collision in near-real-time, and take action. This solution can give light rail drivers a reliable warning system of dangerous situations and provide adequate support to avoid collisions and accidents.
b. Railway intrusion detection. Help drivers detect unusual objects in underground or above-ground tracks. This system would notice crossing objects like vehicles, other trains, or pedestrians on above-ground tracks. This is especially helpful when light conditions are poor due to unground tunnels or weather conditions.
c. Autonomous driving. If the system detects an object too close to provoking an accident, it could send a visual and audio alert to the driver. But if for any reason, the driver does not react within a timeframe (i.e., two seconds), the MV automated system would slow down the light train, eventually stopping or applying emergency breaks, if necessary.
d. Comply with regulations. Collision avoidance systems are mandatory in some countries, especially for fleet vehicles or passenger vehicle transportation such as light rail. Systems that improve the safety of passengers, like Autonomous Emergency Braking (AEB), are, in fact, becoming a standard in carmaking countries like the U.S.
For more information on other edge AI appliances, AI starter kits, or the Machine Vision for light rail collision avoidance, please contact Lanner’s sales representative.
Industrial Grade AI Inference System For 5G Edge With NVIDIA® Jetson NX
|CPU||6-core NVIDIA Carmel ARM®v8.2 64-bit CPU 2MB L2 + 4MB L3|