The adoption of AI is indicative of how the security market is always looking to stay one step ahead of the bad actors it protects against. Using AI is another, hugely powerful tool that it can employ in the battle. However, AI is also complex and processor-intensive. The goal is to deliver better results, faster. AI can do that, at the cost of processing horsepower.
Machine learning models and neural networks are software-based, so of course they can run in standard architectures. But the mathematical complexity of the models puts pressure on the resources of standard processor architectures. AI accelerators are hardware-based subsystems specifically designed to execute complex math faster. For many applications, using an AI accelerator is practically a prerequisite. But these accelerators can also be power-intensive.
Some security systems may run from power supplies that are unstable or have limited capacity, relying on batteries and possibly even solar power for systems in very remote locations. Energy efficiency for any system running 24/7 is important, but this is amplified when using AI, coupled with the need for continuous surveillance.
Power is the critical factor
It’s not surprising that the power-related and power-intensive components contribute considerably to the complexity and total costs of an AI-based video security system. These include the cost of purchasing and the ongoing operational cost.
The semiconductor industry, Avnet Silica and its supplier partners are addressing this. Power is now a critical factor for any application. There is no doubt that the general trend is a net increase in the amount of power we consume. This is balanced by the development of more energy-efficient solutions.
AI is already being seen as contributing to the trend for more power. Fortunately, we are working hard to develop solutions for Edge AI that consume less power. Running machine learning at the edge is being supported with new processor and microcontroller solutions, and the supplier focus is on both performance and power.
At a higher level, the total system power still needs to be managed. As the power requirement continues to rise, manufacturers are looking to deliver more power in the same or even less space. Power density, as it is known, is a challenge, not least because power supplies tend to generate heat. By using more efficient power solutions, the amount of heat (which is wasted energy) can be minimised and the need for forced air cooling (another power-consumption) better managed.
This industry-wide imperative is being addressed with the development of advanced wide bandgap materials for power solutions. The key technologies here are Silicon Carbide (SiC) and Gallium Nitride (GaN). Power semiconductors fabricated in these materials offer benefits over traditional silicon.
Wide bandgap power semiconductors are proven to be more efficient. In applications with high power requirements, moving to SiC or GaN can deliver better energy efficiency and higher power density.
