Road ecologists have proposed a wide range of mitigation measures in order to reduce the number of animal-vehicle collisions (AVC). Some measures aim to modify driver behaviour (road signs, speed limitation, etc.), others to modify animal behaviour (fences, wildlife passes, swareflex, mirrors, etc.), and others both. The costs and effectiveness vary enormously. In general, driver-oriented measures as road signs are widely implemented because of their low cost, but they are also much less effective. This lack of effectiveness is related to driver habituation. Improving sign designs to increase driver response may reduce AVC. To do this, it is possible to take advantage of the fact that AVC are concentrated in time and space. The aim is to alert the driver only when a certain risk threshold is exceeded. With this purpose spatio-temporal AVC models can be used to focus the warning signal on the periods of maximum probability of an accident. The models have to be fed in real time with data about traffic, weather, hunting, road conditions, etc. The results can be displayed on different devices: app, navigators, road signs, etc. For example, we are developing a prototype of variable road sign based on these spatio-temporal AVC models. This prototype will probably show considerable advantages over the measures currently being implemented to minimize AVC. In economic terms, it is much cheaper than other structural measures, and, by focusing only on those moments of real risk, it avoids driver habituation. It is expected that this warning will cause drivers to slow down, so the number of AVC is expected to decrease.The main challenge is to define appropriate alert thresholds in relation to the risk of AVC. A very low threshold would favor driver habituation. A very high threshold would mean that in moments of high risk no alarm would be emitted with the consequent loss of efficiency. The idea could also be applied to endangered species also affected by road traffic (carnivores, amphibians, etc.).