Nighttime driving scenarios are more deadly than daytime. With a much higher fatality rate at night, driving at night is challenging for humans and even more so for self-driving cars.
Driverless technology on the ground needs to meet the requirements for normal vehicle operation under a variety of adverse factors (e.g., extreme environments such as extreme cold, extreme heat, rain, snow, and low-light), with low-light environments being one of the most significant adverse factors, while obstacle recognition such as vehicles is the most important detection item in driverless driving.
As vision-centered autonomous driving systems rely more and more on camera sensors, addressing safety hazards in low-light conditions has become increasingly critical to ensure the overall safety of these vehicles.
One intuitive solution is to collect large amounts of nighttime driving data. However, this approach is not only labor-intensive and costly, but also has the potential to impair the performance of daytime models due to the differences in image distribution between nighttime and daytime.
At night, however, most highway segments can even be categorized as “unlit”, with drivers relying on passive light sources (headlights, reflective pins, moonlight), and more accidents occurring on nighttime curves that are foggy, rainy, or snowy.
Ingletec offers an innovative assisted driving solution for guiding driverless vehicles at night using strobe light sources buried in highway lane dividers. This solution can be realized with the following technology and offers several significant advantages.
Solution:
Vehicle Recognition of Strobe Ground Light Paths
1.1 Visual Sensors
– Camera Detection: Driverless vehicles are equipped with cameras that capture strobe lights on the road. Highly sensitive, low-light cameras can accurately capture changes in the brightness of strobes.
– Image Processing Algorithms: The vehicle’s processing unit analyzes the image frames captured by the camera through image processing algorithms to identify the frequency and location of flashing ground lights. These algorithms typically include:
– Edge Detection and Shape Recognition: Used to extract the position of the floor lamp in the image.
– Timing analysis: to identify flicker patterns by analyzing changes in the brightness of the light source in consecutive frames.
1.2 Optical sensors
– Optical sensors: Vehicles can be equipped with specialized optical sensors for detecting light sources on the road. These sensors are highly sensitive to flickering light sources of specific frequencies and can directly detect signals from strobe ground lights.
1.3 Signal Processing and Sensor Fusion
– Frequency domain analysis: Analyze the frequency characteristics of the light source through Fourier transform and other techniques to accurately identify the specific flashing frequency of the strobe light and distinguish it from other environmental light sources, such as signal light sources, headlights, illuminated signage, and LED road marking lights.
– Multi-sensor fusion: Combine the data from visual sensors, optical sensors and other sensors of the vehicle (e.g. GPS, IMU, etc.) to form an accurate judgment of the vehicle’s position and direction.
Advantages of the program
2.1 Enhance night driving safety
Strobe light sources provide clear lane markings, especially at night when visibility is low or there is no light, and they can significantly improve the lane keeping ability of driverless vehicles. This is critical to preventing vehicles from drifting out of their lanes, especially around sharp bends or on sections of road where visibility is obstructed. This is effective for any form of vehicle.
2.2 Enhanced positioning accuracy
Strobe floor lamps can be used as a kind of “road sign” to help vehicles maintain accurate positioning in the event of weak or lost GPS signals. By continuously recognizing the position of the lights, the vehicle can adjust its position in real time to stay in the right lane. The key is that the positioning accuracy of our product can reach within 10cm.
2.3 Reduce dependence on high-precision maps
By utilizing strobe ground lights on the road, driverless vehicles can reduce their dependence on high-precision maps. This is particularly useful on highways, where real-time changes to the road (e.g., temporary construction, traffic control, etc.) may not be reflected in the map in a timely manner, and the ground lights provide real-time information on road conditions.
2.4 Responding to Adverse Weather Conditions
Strobe ground lights are an active light source that outperforms traditional road markings and reflective materials in foggy days and rainy and snowy weather. As the light can penetrate rain and fog, vehicles can still clearly identify lane lines and maintain safe driving in these conditions.
2.5 Energy efficient and low maintenance
ingletec’s Strobe Ground Light utilizes high-brightness LED technology for low energy consumption and long life. We designed it to power all the modules through high efficiency solar energy storage, and all the components are maintained by low-power power supply, so the system not only has very low maintenance cost, but also doesn’t need to pay high electricity bill, which is very suitable for large-scale deployment on the highway.
2.6 Remote Control for Vehicle-Road Synergy
This solution can be combined with Intelligent Transportation System (ITS) to realize vehicle-road coordination. Ground lights can change their flashing patterns according to real-time road conditions or instructions from the traffic management center, conveying specific light flashing guidelines or warning messages to the driverless vehicle, and on-board computation can recognize the frequency codes based on a preset model to instantly execute evasions.
Potential Challenges
Despite the many advantages of this solution, there are some challenges:
– Initial Deployment Costs: Large-scale installation of strobe ground lights may require a high initial investment, which needs to be recognized by traffic authorities as it will improve the safety of all highway travelers and is not limited to driverless vehicles.
– Data fusion complexity: Multi-sensor data fusion and processing requires high computational resources and optimized algorithms, but with the cost of computing power and optimization of algorithms, a balance will be reached in the near future.
