Algorithm 1: Q-learning-based Backoff Mechanism
1	Initialization:
2	NCCA = 0, load_Q_table() // load a previously learned value
3	Input: medium channel state by CCA (idle or busy), packet transmission (success of failure)
4	Output: backoff duration (backofftime)
5	while mac_protocol do
6	if active_mode
7	busy_count += CCA(is_busy) or busy_count −= CCA(is_idle)
8	// channel mode is set to 4 levels according to busy_count
9	ch_mode = channel_mode(busy_count);
10	end if
11	if event_send_request
12	while NCCA < macMaxCSMABackoffs or packet_transmitted
13	// set the largest value in the Q table to next CW
14	CWt = Max_Q_table(ch_mode, CWt − 1)
15	Backoff(backofftime = uniform_random(0, 2CWt − 1) )
16	if CCA(is_busy)
17	busy_count += 1; NCCA += 1;
18	else if CCA(is_idle)
19	transmit(packet) and wait(Ack)
20	busy_count −= 1; NCCA = 0;
21	end if
22	ch_mode = channel_mode(busy_count)
23	if learning_mode
24	Q(st[ch_mode]: CWt − 1, at: CWt) - Equation(2)
25	end if
26	// network parameter measurement
27	transmission_success_ratio ($r_t$) = # of received_ack / # of requested packet,
28	average_latency ($l_a$) = ack_time – transmit_time (about only transmitted packet)
29	// other parameters else (e.g., meanBackoffduration, channelAccessratio, etc.)
30	end while
31	delete_queue(packet) // transmitted or dropped
32	end if
33	end while