"It can see, hear, feel, and smell and it will roam around autonomously 24/7," said CEO William Santana Li, a former Ford Motor executive, in an interview with CNET.
At the moment, the K5 is only a prototype, and Knightscope next year will launch a beta program with select partners. But the company is shooting to have the K5 fully deployed by 2015 on a machine-as-a-service business model, meaning clients would pay by the hour for a monthly bill, based on 40-hour weeks, of $1,000. The hourly rate of $6.25 means the cost of the K5 would be competitive with the wages of many a low-wage human security guard.
Servicing and monitoring of the bots will depend on client needs, Li said, with either Knightscope or the customer employing someone to manage the bots full-time.
Crime prediction is one of the more eye-popping features of the K5, but the bot is also packed to the gills with cutting-edge surveillance technology. It has LIDAR mapping -- a technique using lasers to analyze reflected light -- to aid its autonomous movement. "It takes in data from a 3D real-time map that it creates and combines that with differential GPS and some proximity sensors and does a probabilistic analysis to figure out exactly where it should be going on its own," Li explained.
It also has behavioral analysis capabilities and enough camera, audio, and other sensor technology to pump out 90 terabytes of data a year per unit. Down the line, the K5 will be equipped with facial recognition and even the ability to sniff out emanations from chemical and biological weapons, as well as airborne pathogens. It will be able to travel up to 18 mph, and later models will include the ability to maneuver curbs and other terrain.
The K5 will not be armed. Still, teens with late-night bot-tipping ambitions had best beware, lest their hijinks be recorded for posterity, and possible prosecution. Li said that messing with a Knightscope bot -- which would be difficult given its weight -- will have serious ramifications, as would tampering with any other form of security equipment on private property.
Still, the most sci-fi of all its features, the crime prediction algorithms, do sound too good to be true. And to be more precise, the K5 won't be so much predicting crime as much as it will be analyzing multiple data points simultaneously and knowing when a situation may be on the precipice of becoming dangerous.
"Predicting crime is being deployed today, but it's unfortunately using a lot of historical data," Li explained. "What doesn't exist in that algorithm is real-time on-site data. So if you actually had data that was fresh, that was actually from the location you're trying to analyze, it would make that algorithm much more robust." Li noted that the main goal of the crime prediction algorithms and autonomous function is to be able to push out an alert early with that kind of data, as well as aid the K5 in knowing when to charge itself and what time of day or night is optimal for uploading and downloading data in a specific environment.
