Tiny, Distributed, and Eco-Optimized: Proposal of Design Guidelines for Environmentally Friendly ML Devices

The researchers in the DESIRE4EU project propose the use of an eco-optimized PLA/Flax-based PCB manufacturing substrate that could be recycled or degraded after use due to the organic, biodegradable nature of the material. However, having a biodegradable PCB is not enough, as we need to rethink electronics design to better fit specific use cases, avoiding a one-size-fits-all philosophy. This paper includes a set of learned lessons in the form of design guidelines extracted from the experience of a small-scale manufacturing of a set of biodegradable microcontroller boards. These lessons outline contemporary technical limitations of biodegradable PCBs, which the authors trust will be overcome during the development of the mentioned EU project by 2027. This should be a first step towards reducing e-waste in the not-so-far future. This positioning paper states that, in the current ecological crisis, the different engineering communities need to rethink their priorities in order to produce ecology-friendly innovations by keeping concepts such as just enough computing and ecooptimization in mind. In particular, when talking about the TinyML community, we suggest carefully considering the hardware's limiting factors presented by computational power, or radio communication when designing new Edge devices so that they could use 2-layers biodegradable PCBs. In parallel to the hardware discussion, the authors bring up issues emerging from using bloated inference software production workflows, which have a very direct impact in the ecology due to the computation power needed for embedded machine learning software production. This paper suggests the adoption of ondevice training to minimize the energy consumption and dependance on connected toolchains during programming.