Imagine a team of autonomous drones, armed with advanced detection equipment, looking for smoke as they fly over the Sierra Nevada mountains. As soon as they discover a forest fire, these leading robots send instructions to a swarm of fire fighting drones, which are quickly on their way to the scene of the fire.
But what if one or more lead robots are hacked by a malicious agent and send wrong instructions? If the follower robots move away from the fire, how do they know they’ve been tricked?
Using blockchain technology as a communication tool for a robot team could provide security and protection against deception, according to a study published today in IEEE Transactions on Robotics by researchers at MIT and the Polytechnic University of Madrid. It can also find application in cities, where self-driving multi-robot systems for cars deliver goods and move people around the city.
A blockchain offers a tamper-proof recording of all transactions, in this case the messages from the robot team leaders, so that the follower robots can recognize any inconsistencies in the information path. Executives use tokens to signal movement and add transactions to the chain, and they lose their tokens if caught lying, so this transaction-based communication system limits the number of lies a hacked robot could spread, according to Eduardo. Marie Curie Fellow at the MIT Media Lab and lead author of the article.
Not Just for Bitcoin
While a blockchain is generally used as a secure general ledger for cryptocurrencies, it is essentially a list of data structures called blocks that are linked in a chain, each block containing information that needs to be stored, the “hash” of the information in the block and the “hash” of the previous block in the chain. Hashing is the process of converting a text string into a series of unique letters and numbers.
In this simulation-based study, the information stored in each block is a series of addresses from a leading robot to its followers. When a malicious robot tries to change the contents of a block, it changes the hash of the block so that the changed block changes. No longer connected to the chain, changed addresses could easily be ignored by follower bots.
The blockchain also provides a permanent record of all transactions. Finally, since all followers can see all the instructions from the leading robots, they can see if they have been misled.
For example, if five leaders send messages telling their followers to move north and one leader sends a message telling them to move west, followers could ignore this inconsistent direction. Even if a follower robot accidentally moved west, the deluded robot would eventually spot the error by comparing its movements with transactions stored on the blockchain.
Transaction-based communication
In the system developed by the researchers, each leader is given a fixed number of tokens that are used to add transactions to the chain: a token is needed to add a transaction when followers discover that the information in a block is false, by checking what most of the leading robots have shown in that particular step, the leader will lose the token. Once a robot runs out of tokens, it can no longer send messages.
The researchers tested their system by simulating various leader tracking situations where the number of malicious bots was known or unknown. Using a blockchain, executives sent instructions to follower robots moving through a Cartesian plane, while malicious executives sent false instructions or tried to block the path of follower robots.
Researchers found that even when follower bots were initially fooled by malicious leaders, the transaction-based system allowed all followers to eventually achieve their goal. And because each leader has an equal finite number of tokens, the researchers developed algorithms to determine the maximum number of lies a malicious robot can count.
Since, it was discovered how lies can affect the system and a malicious robot can inflict damage on the system, they were able to calculate the upper limit for deceiving the swarm. If you have robots with a certain battery life, it doesn’t matter who hacks the system, the robots have enough battery to achieve their goal.
The algorithms allow a system designer not only to calculate the battery life the robots will need to do their job, but also the amount of memory it will take to store the blockchain, the number of robots it will need, and which way it will take them can leave even if a certain percentage of the leading robots are hacked and become malicious.
In the future, Castelló hopes to use this work to create new safety systems for robots using transaction-based interactions, and sees this as a way of building trust between humans and groups of robots.
If you turn these robotic systems into public robotic infrastructure, you expose them to malicious actors and errors. These techniques are useful in validating, testing, and understanding that the system is not becoming dishonest. If members of the system are hacked, the infrastructure will not be destroyed, Castelló added.
The work was co-authored by Ernesto Jiménez and José Luis López Presa from the Polytechnic University of Madrid and the Horizon 2020 Research and Innovation Program of the European Union, the Community of Madrid and the MIT International Science and Technology Initiatives Global Seed.
