CHIRP
Most IoT networking assumes resources the smallest devices do not have. A conventional stack expects memory, a scheduler, and a power budget that a sensor node running for years on a coin cell cannot supply. The usual answer is to centralize — push the intelligence to a gateway and leave the endpoint dumb — which trades away autonomy and creates the single point of failure the deployment was often trying to avoid.
CHIRP implements a peer-to-peer network using the CHIRP communication algorithm, targeting exactly that constrained regime. The research question is not whether the algorithm works in principle; it is what the algorithm costs once it stops being a specification and becomes firmware on a real radio.
Concretely, we are measuring:
- Power, per message and at idle, since idle dominates the lifetime of anything battery-powered.
- Memory, both code and state, against the footprint of parts that are actually cheap in volume.
- Latency and delivery, under contention and interference rather than on a clean bench.
The lab’s earlier S-CHIRP work addressed the security layer over this protocol family. Establishing the unsecured baseline honestly is a prerequisite to saying what security costs on top of it.
Open questions
- What is the true idle power of a peer-to-peer node, and does it undercut a gateway-based deployment over a realistic device lifetime?
- How does delivery degrade as node density rises, and where does the protocol stop being peer-to-peer in practice?
- What headroom is left for security once the base protocol has taken its share of a constrained part?
Aegis Next project
Dark Logic