Overview

In the TickTalk (TT) project (see our project page and repository), our objective is to make easier the problem of programming large arrays of loosely-timed systems, such as the sensors in a smart city. In these systems:

  • Time matters

  • Parallelism is inherent

  • Hardware elements are heterogeneous

  • Power may be scarce

The programmer will be forced to deal with these issues, and improvements in one domain may adversely impact another, like higher precison time synchronization at the cost of power.

Our fundamental hypothesis is that programming becomes easier when it is possible to separate the questions of “Is this program logically correct?” from “How do I optimally map this problem onto a large, loosely-timed array of possibly heterogeneous computing elements?”

_images/WhyIsTimeImportant.jpg

Focusing on programming with time, we observe that it is fundamental to many applications of distributed time-sensitive applications in Cyber Physical Systems (CPS) and IoT. We care about when physical quantities are sampled, we care about being able to fuse multiple sensor readings that were taken at the same time, and we care about being able to actuate things in the real world (e.g., traffic signals) at specified times. Historically, doing these things in a programming language necessitated stepping outside of the language itself because time was not represented as something foundational to the language. With TickTalk, we think this limitation should be overcome, and in fact the project name – TickTalk – derives from the idea that programming languages should enable programmers to express (or talk about) notions of time (expressed in ticks).

TTPython is one such programming language. Built on the base of Python3, TTPython seeks to add language constructs that enable programmers to express timing concepts from the perspective of what to do with time rather than how to do it. TTPython was conceived in January of 2021 as a language that can be used within the TickTalk project to support research as well as a language that can help to broaden the discussion about time-based programming for loosely-timed systems. The latter is an important point. We are not seeking a language that meets the requirements of hard real-time systems. We are not seeking a language that helps assure worst-case execution times are met. There is extensive work that has been done by others that covers these topics.

Instead, we enable non-specialist programmers to write programs for potentially massive IoT sensor and actuator networks, in which statistical precision is sufficient and in which time accuracy at a large scale (many sensors being aggregated instead of depending on handling the data from just one) is more important than precision at a small scale. TTPython enables programmers to write distributed, time-sensitive applications as a singular program, in which the program’s temporal and mapping, i.e., on which devices some piece of code runs, can be specified in one place without requiring the programmer to extensively describe interfaces and protocols among their distributed, heterogeneous system components.