Abstract:
There is a huge demand in the field of Artificial intelligence for making novel works that
can mimic human creativity. One interesting example is handwritten calligraphy. Robotic
calligraphy, as a typical application of robot movement planning, is of great significance
for the education of calligraphy culture. The existing implementations of such robots
often suffer from their limited ability for font generation and evaluation, leading to poor
writing style diversity and writing quality. These demands are increasing rapidly, hoping
to create something intelligent enough to pace with them at handwritten calligraphy
without human intervention. The core idea of our project is to provide a solution that
humans can utilize, to mimic the calligraphy of handwritten texts using a robotic system.
The robotic system can learn from previous artworks, and produce similar strokes with a
sense of learning creativity. In our work, we aim to reach a satisfying level where
handwriting is done automatically, creatively, and with astonishing results. The work is
utilizing reinforcement learning techniques with a long-short-term memory (LSTM)
model alongside a generative adversarial model (GAN to be used as a proof of concept)
to achieve an approach in which the robot will be guided. This work has shown by its
outcomes that it’s possible to write words smoothly on physical surfaces using the robot
and not only soft-writing but as well to mimic a specified human writing style from
images after pretraining the system on sequences of points of lines belonging to other
styles
