Group proposes AI system to predict, prevent roadkill

TaiRON's proposal outlines a two-pronged approach to mitigate wildlife road deaths and associated traffic incidents. The first component involves an AI system that integrates various datasets, including TaiRON's own records of animal deaths, wildlife activity data from the Taiwan Biodiversity Network, and official traffic accident statistics. This data, cross-referenced with environmental factors like road maps, speed limits, traffic volume, microclimate data, and Google Street View images, would allow the AI to simulate and predict specific areas where conditions are frequently associated with roadkill.

This predictive AI aims to shift prevention from reactive sign placement to proactive identification of danger zones. Initially, the system would focus on pangolins as a pilot species, with plans to expand its scope to include other vulnerable wildlife such as leopard cats and endangered land crabs. The output from this data-driven prediction could inform where resources are best deployed for prevention, rather than waiting for incidents to occur.

The second component is a real-time early warning network. TaiRON proposes installing electronic signs equipped with heat and light sensors to detect animals approaching or crossing roads, thereby alerting drivers. The group has developed a low-cost, solar-powered warning device, which uses infrared sensors for warm-blooded animals and light sensors for vehicles, designed for operation in low-light conditions. This device is slated for testing on roads managed by the New Taipei City Government before potential nationwide deployment.