When a biker looks for a new adventure by picking a long gone trail after the rainy season, he expects fresh amendments in his maneuvering, meanwhile to assist him our old-gen bike learns new vents to drip engine-oil. Before asserting “not anymore” we must agree with fag complexity of almost pure mechanical propulsion system and methods to make it learn from pattern. Even for EVs, it comes as a challenge to tune power factors concurrently just relying on current parameters from a wide range of sensors. Every algorithm which tries to manipulate driving components just by mathematical equation between domains would make a vehicle less responsive to challenges and vice-versa. Solving this with averaging out values from passed time-stamps gives a new set of problems such as eased-out functions and then everything, for overstating “Hitting bull’s eye with dependent values and a static algorithm is okay! unless algorithm feels like a one-eyed cat looking into two holes.”
Starting with an approached solution, if your pet algorithm is humble enough to evolve and learn from the decisions made yesterday, I would like to call it a day. But worst-case scenarios are the benchmarks for any test-case algorithms and I don’t hope a story of dystopia for every single vehicle just to make it ready for it, now every bike can be divided into groups by the independent parameters such as regions, terrain variations, temperature ranges, riding style and etc, these bikes can do regression on their respective cases to achieve a certain point in ever-evolving algorithm and then these milestones can be shared across the sister-machines from the same region.
Hitherto one question should have haunted us “Is device side hardware is capable enough to do plotting and processing at such a level, what about server?” This is why we have loaded our hardware with 1.3 GHz application processor, running a custom OS utilizing it, alongside on the processor we have mounted every sensor which is required to have extrinsic parameters of the vehicle. For the server side, it has to be equipped with appropriate hardware as granted but what does distinguish them to maintain the efficiency is an efficient algorithm which lets him maintain a good pool of connections concurrently. This data and processing house is having sophisticated lines of data compression, concurrent connections real-time availability. We have fed both compression and security to the same network layer to avoid heavy parsing which is a good boost in concurrency. Block transfer with dynamic size helps us to solve server spamming from too lite frames and then big frames don’t get bigger and bigger till the size of charging Rhino.
Now since we have tamed our network communication, the next focus is on real-time plotting. We as a powerpoint generation are very well versed in understanding the number of charts, but an electric drive train has so many of dependable intrinsic parameters which need a good interpolation method to know the anomalies in the system just by looking at it, it should be intuitive and adaptable, mentioning good-looking at last. By experience we have switched to 3D-graph for modern browsers, having separate methods for both MAP-based plotting and parameter-based plotting, we found Web-GL suitable for the task and we have optimized it to suit native graphic-capability of the processor along with dedicated GPUs. The major edge of 3D plotting is that it provides intuitive interpolation which does not look crowded with heavy data set even after zooming out. For more yield time at the server side, we have pushed data de-framing at client APIs so it slightly increases the code base for browser side but shows many improvements it data-base handling and processing.
Whatever we have mentioned yet is part of put-together software architecture which we call as TIROS, such data filtration, handling and plotting methods are base to our debugging and machine learning bench, where it is yet evolving to cover service and a dealership based problems.
Coding machines gives us a chance to learn best about ourselves.
while ( life )
learn( life ) ;